Add files using upload-large-folder tool

marked/P/T-REC-P.1199-202510-I_PDF-E/raw.md

@@ -0,0 +1,720 @@
+
+
+# Recommendation
+
+## **ITU-T P.1199 (10/2025)**
+
+SERIES P: Telephone transmission quality, telephone installations, local line networks
+
+Communications involving vehicles
+
+---
+
+## **Parametric object-recognition-ratio-estimation model for remote monitoring of surveillance video delivered from autonomous vehicles**
+
+![ITU logo](84a1d09fb489061482111515543b60dc_img.jpg)
+
+The logo of the International Telecommunication Union (ITU) is located in the bottom right corner. It features a blue globe with white lines representing latitude and longitude, and the letters 'ITU' in a bold, blue, sans-serif font superimposed on the globe.
+
+ITU logo
+
+## ITU-T P-SERIES RECOMMENDATIONS
+
+## Telephone transmission quality, telephone installations, local line networks
+
+|                                                                                                    |                      |
+|----------------------------------------------------------------------------------------------------|----------------------|
+| Vocabulary and effects of transmission parameters on customer opinion of transmission quality      | P.10-P.19            |
+| Voice terminal characteristics                                                                     | P.30-P.39            |
+| Reference systems                                                                                  | P.40-P.49            |
+| Objective measuring apparatus                                                                      | P.50-P.59            |
+| Objective electro-acoustical measurements                                                          | P.60-P.69            |
+| Measurements related to speech loudness                                                            | P.70-P.79            |
+| Methods for objective and subjective assessment of speech quality                                  | P.80-P.89            |
+| Voice terminal characteristics                                                                     | P.300-P.399          |
+| Objective measuring apparatus                                                                      | P.500-P.599          |
+| Measurements related to speech loudness                                                            | P.700-P.709          |
+| Methods for objective and subjective assessment of speech and video quality                        | P.800-P.899          |
+| Audiovisual quality in multimedia services                                                         | P.900-P.999          |
+| Transmission performance and QoS aspects of IP end-points                                          | P.1000-P.1099        |
+| <b>Communications involving vehicles</b>                                                           | <b>P.1100-P.1199</b> |
+| Models and tools for quality assessment of streamed media                                          | P.1200-P.1299        |
+| Telemeeting assessment                                                                             | P.1300-P.1399        |
+| Statistical analysis, evaluation and reporting guidelines of quality measurements                  | P.1400-P.1499        |
+| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500-P.1599        |
+
+For further details, please refer to the list of ITU-T Recommendations.
+
+# Recommendation ITU-T P.1199
+
+# Parametric object-recognition-ratio-estimation model for remote monitoring of surveillance video delivered from autonomous vehicles
+
+## Summary
+
+Recommendation ITU-T P.1199 provides a parametric object-recognition-ratio-estimation model to check whether an observer in a monitoring centre can recognize an object (e.g., a person who jumps out onto the road or debris on the road) while viewing the video taken by an autonomous vehicle's surveillance camera and delivered to the monitoring centre. This object-recognition ratio can be used as an indicator to check whether the surveillance video delivered to the remote monitoring centre is of sufficient quality for remote monitoring around the autonomous vehicle.
+
+The input used by the model consists of information obtained from video streams and vehicle information. Four different modes, which are called modes of operation, can be used for estimating the object-recognition ratio in this Recommendation:
+
+- Mode 0: Information obtained from the video stream, such as the video resolution, video bitrate and video frame rate, in addition to packet loss information.
+- Mode 1: Information obtained from the video stream and frozen video frame information based on frame inspection.
+- Mode 2: The same information as Mode 0, with the addition of vehicle velocity.
+- Mode 3: The same information as Mode 1, with the addition of vehicle velocity.
+
+## History\*
+
+| Edition | Recommendation | Approval   | Study Group | Unique ID          |
+|---------|----------------|------------|-------------|--------------------|
+| 1.0     | ITU-T P.1199   | 2025-10-29 | 12          | 11.1002/1000/16486 |
+
+## Keywords
+
+Autonomous driving, object recognition, surveillance video.
+
+---
+
+\* To access the Recommendation, type the URL <https://handle.itu.int/> in the address field of your web browser, followed by the Recommendation's unique ID.
+
+## FOREWORD
+
+The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, and information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis.
+
+The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics.
+
+The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1.
+
+In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC.
+
+## NOTE
+
+In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency.
+
+Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party.
+
+## INTELLECTUAL PROPERTY RIGHTS
+
+ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process.
+
+As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at <https://www.itu.int/ITU-T/ipr/>.
+
+© ITU 2026
+
+All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU.
+
+## Table of Contents
+
+|                                                                                  | Page |
+|----------------------------------------------------------------------------------|------|
+| 1 Scope.....                                                                     | 1    |
+| 2 References.....                                                                | 2    |
+| 3 Definitions .....                                                              | 2    |
+| 3.1 Terms defined elsewhere .....                                                | 2    |
+| 3.2 Terms defined in this Recommendation.....                                    | 2    |
+| 4 Abbreviations and acronyms .....                                               | 2    |
+| 5 Conventions .....                                                              | 3    |
+| 6 Monitoring point.....                                                          | 3    |
+| 7 Areas of application.....                                                      | 3    |
+| 7.1 Applications for which the model is intended.....                            | 3    |
+| 7.2 General application range to which the model is applicable.....              | 3    |
+| 7.3 Modes of operation.....                                                      | 4    |
+| 8 Building blocks.....                                                           | 4    |
+| 8.1 Model inputs.....                                                            | 5    |
+| 8.2 Model outputs.....                                                           | 7    |
+| 9 The model algorithm.....                                                       | 7    |
+| Annex A – Calculation of target distance .....                                   | 11   |
+| Appendix I – Performance figures .....                                           | 13   |
+| Appendix II – Camera-related factors.....                                        | 15   |
+| Appendix III – Guidelines for the use case of the object-recognition ratio ..... | 16   |
+| III.1 Use Case 1: Alerting observers at the monitoring centre .....              | 16   |
+| III.2 Use Case 2: Detecting object-recognition-ratio-degradation areas .....     | 18   |
+| Bibliography.....                                                                | 19   |
+
+
+
+# **Parametric object-recognition-ratio-estimation model for remote monitoring of surveillance video delivered from autonomous vehicles**
+
+# **1 Scope**
+
+This Recommendation provides a parametric object-recognition-ratio-estimation model to check whether an observer in a monitoring centre can recognize an object (e.g., a person who jumps out onto the road or debris on the road) when viewing the video taken by an autonomous vehicle's surveillance camera and delivered encoded video to a monitoring centre (hereafter, the video viewed at this monitoring centre is referred to as surveillance video).
+
+No driver is needed in fully autonomous driving, and the autonomous vehicle operates autonomously in any location, on any road and in any weather. For safe driving, objects that interfere with driving need to be automatically detected, for example, to brake and decelerate the vehicle. Therefore, an autonomous vehicle is equipped with object-detection technology [b-Cai] to detect objects around the autonomous vehicle by analysing videos captured by the automotive cameras and LiDAR data. Under certain conditions, such as on a highway, object-detection technology in autonomous driving works well. On the other hand, current object-detection technology is challenging to use on local streets because various objects appear, such as traffic signs or pedestrians.
+
+In several countries, remote monitoring is legally required for autonomous driving to ensure safety. In Germany, legislation on autonomous driving has been passed that states that remote monitoring is necessary [b-German Gov.]. Also, under the revised Road Traffic Act in Japan, remote monitoring is needed when operating unattended autonomous vehicles in specific areas [b-Ikeuchi]. In France, the revised transport law requires that the safety of autonomous driving is ensured through remote monitoring [b-French Gov.]. On the basis of the above, to check whether remote monitoring is being carried out properly, it is important to check whether an observer at the monitoring centre can recognize an object using the surveillance video.
+
+However, since the surveillance video is encoded and delivered through a network, video quality may degrade as the bandwidth of the wireless network decreases, making it difficult for observers to recognize objects in the video. It is a meticulous and difficult task for the observer to continuously judge whether the object can be recognized from surveillance video in real time. Furthermore, in such a case, the observer needs to react and brake the autonomous vehicle immediately when an object appears, so what is necessary is to confirm that an object appears in the surveillance video rather than to identify the object type. Therefore, to check whether remote monitoring is being carried out properly, a parametric object-recognition-ratio estimation model needs to be developed to estimate the object-recognition ratio, which is the percentage of observers who can recognize objects within the time it is possible to prevent the vehicle from colliding with them when these objects appear a target distance away from the autonomous vehicle.
+
+There is a wide range of cases where object recognition is required for remote monitoring in autonomous driving, e.g., there may be objects on the road or pedestrians jumping out onto the road. Since a model corresponding to all these situations is difficult to construct, a model needs to be built that considers the riskiest situations for autonomous driving. From the viewpoint of the risk of a serious accident and the difficulty in recognition (i.e., the need for instantaneous judgment by an observer), the case in which an object appears in front of the vehicle is covered in this Recommendation.
+
+The object-recognition ratio is affected by various factors. The video may be degraded due to the video bitrate reduction and the packet loss, which makes object recognition difficult. The network latency affects the object-recognition ratio because if the network latency is long, the observer must recognize objects earlier to react. The vehicle's velocity also affects the object-recognition ratio
+
+because if the vehicle's velocity is high, the braking distance becomes longer, and the observer must recognize the object far from away when it appears in front of the vehicle. The camera specifications (e.g., focal length and field of view) and problems (e.g., dirt on the lens) also affect the object-recognition ratio. Therefore, a parametric object-recognition-ratio estimation model takes the encoding, packet loss, and velocity information as input and uses a coefficient set determined by other factors (e.g., object, camera, and weather) on the basis of a priori information. Note that since depending on the configuration of the remote monitoring system, since the velocity of the autonomous vehicle and the effect of packet loss on the video (i.e., video-frame-loss information) may not be possible to obtain, the model has four modes depending on the available input information.
+
+# 2 References
+
+The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation.
+
+None.
+
+# 3 Definitions
+
+## 3.1 Terms defined elsewhere
+
+This Recommendation uses the following term defined elsewhere:
+
+**3.1.1 timed task method** [b-ITU-T P.912]: A viewer is asked to watch for a particular action or object to be recognized in the video clip.
+
+## 3.2 Terms defined in this Recommendation
+
+This Recommendation defines the following terms:
+
+**3.2.1 control latency:** The time between an observer sending a brake command and the autonomous vehicle starting to stop.
+
+**3.2.2 glass-to-glass latency:** The sum of delays associated with network, encoding, decoding and display latency.
+
+**3.2.3 required reaction time:** The time within which an observer must recognize an object in the path of an autonomous vehicle and react.
+
+**3.2.4 target distance:** A safety threshold in autonomous vehicle operation defining the minimum distance between the front of a vehicle and an object appearing in its path at which, after accounting for all system and human delays and the vehicle's braking distance, the vehicle can be stopped safely before collision.
+
+NOTE – Human delays include glass to glass latency, observer reaction time and control latency.
+
+# 4 Abbreviations and acronyms
+
+This Recommendation uses the following abbreviations and acronyms:
+
+GOP Group of Pictures
+
+HEVC High-Efficiency Video Coding
+
+IP Internet Protocol
+
+|      |                                 |
+|------|---------------------------------|
+| PCC  | Pearson Correlation Coefficient |
+| RMSE | Root Mean Square Error          |
+| RTP  | Real-time Transport Protocol    |
+| UDP  | User Datagram Protocol          |
+
+# 5 Conventions
+
+None.
+
+# 6 Monitoring point
+
+To check whether an observer at the monitoring centre can recognize an object on the surveillance video, the parametric object-recognition-ratio estimation model is implemented at the monitoring centre, as shown in Figure 1. For example, an alert is triggered for an observer to detect video degradation based on a threshold of object-recognition ratio (e.g., 80%) and the observer should react and brake the autonomous vehicle immediately.
+
+![Figure 1 – Monitoring point. A diagram showing the flow of data from an autonomous vehicle to a monitoring centre. On the left, an autonomous vehicle with a surveillance camera and encoder sends 'Video packets' through a 'Wireless network' to a 'Decoder' at the 'Monitoring centre'. The 'Decoder' sends 'Surveillance video' to a 'Surveillance video monitor' and 'Input parameter' to an 'Object-recognition ratio calculation device'. The 'Object-recognition ratio calculation device' also receives 'Vehicle's velocity (Mode 2 and 3)' from the vehicle. It outputs 'Object-recognition ratio per unit time' to an 'Object-recognition ratio monitor'. The monitor displays a graph with a 'Threshold (e.g., 80%)'. An 'Observer' is shown reacting to the monitor. Two text boxes indicate: 'Observer shall react and brake the autonomous vehicle immediately if an alert is triggered' and 'An alert is triggered for an observer if the object-recognition ratio decreases under threshold'.](367926125450c2bc3f4bdca9d59a62ba_img.jpg)
+
+Figure 1 – Monitoring point. A diagram showing the flow of data from an autonomous vehicle to a monitoring centre. On the left, an autonomous vehicle with a surveillance camera and encoder sends 'Video packets' through a 'Wireless network' to a 'Decoder' at the 'Monitoring centre'. The 'Decoder' sends 'Surveillance video' to a 'Surveillance video monitor' and 'Input parameter' to an 'Object-recognition ratio calculation device'. The 'Object-recognition ratio calculation device' also receives 'Vehicle's velocity (Mode 2 and 3)' from the vehicle. It outputs 'Object-recognition ratio per unit time' to an 'Object-recognition ratio monitor'. The monitor displays a graph with a 'Threshold (e.g., 80%)'. An 'Observer' is shown reacting to the monitor. Two text boxes indicate: 'Observer shall react and brake the autonomous vehicle immediately if an alert is triggered' and 'An alert is triggered for an observer if the object-recognition ratio decreases under threshold'.
+
+P.1199(25)
+
+Figure 1 – Monitoring point
+
+# 7 Areas of application
+
+## 7.1 Applications for which the model is intended
+
+The application area for this Recommendation is a remote monitoring system in which surveillance video encoded in an autonomous vehicle is delivered to the monitoring centre. A remote monitoring system is based on the RTP/UDP/IP and UDP/IP protocols because low latency is needed for surveillance video, and UDP-based streaming is widely used for real-time streaming.
+
+## 7.2 General application range to which the model is applicable
+
+The application range for which the model has been validated is detailed in Table 1.
+
+**Table 1 – Application range for which the model has been validated**
+
+|                                                                                                                                    |                                                             |           |  |  |
+|------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------|-----------|--|--|
+| Display resolution                                                                                                                 | 1920 × 1080 pels                                            |           |  |  |
+| Unit time for video and network information acquisition                                                                            | 1 second                                                    |           |  |  |
+| Required reaction time                                                                                                             | 1 second                                                    |           |  |  |
+| Glass-to-glass latency and control latency                                                                                         | Total 1.5 seconds                                           |           |  |  |
+| Timeframe                                                                                                                          | Daytime                                                     | Nighttime |  |  |
+| Weather                                                                                                                            | Sunny                                                       |           |  |  |
+| Road surface coefficient of friction                                                                                               | 0.7 (Note 1)                                                |           |  |  |
+| Object colour and size                                                                                                             | Children (about 120 cm tall) wearing white clothes          |           |  |  |
+| Autonomous vehicle's velocity                                                                                                      | 10-40 km/h                                                  |           |  |  |
+| Packet loss error concealment                                                                                                      | Freezing with skipping                                      |           |  |  |
+| Packet loss rate                                                                                                                   | 0-10%                                                       |           |  |  |
+| Frame loss rate                                                                                                                    | 0-95%                                                       |           |  |  |
+| Video codec                                                                                                                        | H.265/HEVC                                                  |           |  |  |
+| Profile                                                                                                                            | Main                                                        |           |  |  |
+| Group of pictures (GOP)                                                                                                            | All-intra/IPP                                               |           |  |  |
+| Video resolution                                                                                                                   | 640 × 360 (230 400) – 1920 × 1080 (2 073 600) pels (Note 2) |           |  |  |
+| Video bitrate                                                                                                                      | 200–5 000 kbit/s                                            |           |  |  |
+| Video frame rate                                                                                                                   | 10–60 fps                                                   |           |  |  |
+| NOTE 1 – This parameter depends on the road surface and weather, i.e., the coefficient of friction is smaller in the case of rain. |                                                             |           |  |  |
+| NOTE 2 – Video resolution is defined as the number of pels (width × height).                                                       |                                                             |           |  |  |
+
+## 7.3 Modes of operation
+
+The model has four modes of operation, which are defined in Table 2. It is recommended to use Mode 3 if possible. However, if velocity or video-frame-loss information is not available, Modes 0 to 2 will be used. Detailed input information of each mode is provided in clause 8.1.
+
+**Table 2 – The model modes of operation**
+
+| Mode | Input                                                                       |
+|------|-----------------------------------------------------------------------------|
+| 0    | Metadata and packet loss information                                        |
+| 1    | Metadata and frozen video frame information                                 |
+| 2    | Metadata, packet loss information and autonomous vehicle's velocity         |
+| 3    | Metadata, frozen video frame information, and autonomous vehicle's velocity |
+
+# 8 Building blocks
+
+The building blocks of the model are depicted in Figure 2. This Recommendation uses codec-related factors such as video bitrate, network-related factors such as packet loss rate, and vehicle-related
+
+factors such as the vehicle's velocity as inputs. In addition, the coefficient is determined by utilizing *a priori* information such as glass-to-glass latency, weather and timeframe (i.e., daytime or nighttime). The details of this input are given in clause 8.1.
+
+![Figure 2 – Building blocks of the model. This block diagram shows the flow of information from inputs to an output through various processing modules. On the left, 'Stream I.01' and 'Vehicle information I.02' enter an 'Input-level' block. 'Stream I.01' is processed by 'Codec-related factors per unit time' (outputting I.11), 'Network-related factors per unit time' (outputting I.12), and 'Vehicle-related factors per unit time' (outputting I.13). 'Vehicle information I.02' is also processed by 'Vehicle-related factors per unit time'. These three factors (I.11, I.12, I.13) are then fed into an 'Object-recognition-ratio-estimation module'. I.11 provides 'Video resolution', 'Video bitrate', and 'Video frame rate'. I.12 provides 'Packet loss rate' and 'The number of lost frames'. I.13 provides 'Velocity'. Below the input-level, 'A-priori information' (labeled I.GEN) is processed, providing 'Glass-to-glass latency', 'timeframe', and 'weather, etc.' to a 'Coefficients table' within the estimation module. The 'Object-recognition-ratio-estimation module' produces the final output 'O.11'. A reference label 'P.1199(25)' is in the bottom right.](5b4e774d63e0e0ed73801a9247755e5f_img.jpg)
+
+Figure 2 – Building blocks of the model. This block diagram shows the flow of information from inputs to an output through various processing modules. On the left, 'Stream I.01' and 'Vehicle information I.02' enter an 'Input-level' block. 'Stream I.01' is processed by 'Codec-related factors per unit time' (outputting I.11), 'Network-related factors per unit time' (outputting I.12), and 'Vehicle-related factors per unit time' (outputting I.13). 'Vehicle information I.02' is also processed by 'Vehicle-related factors per unit time'. These three factors (I.11, I.12, I.13) are then fed into an 'Object-recognition-ratio-estimation module'. I.11 provides 'Video resolution', 'Video bitrate', and 'Video frame rate'. I.12 provides 'Packet loss rate' and 'The number of lost frames'. I.13 provides 'Velocity'. Below the input-level, 'A-priori information' (labeled I.GEN) is processed, providing 'Glass-to-glass latency', 'timeframe', and 'weather, etc.' to a 'Coefficients table' within the estimation module. The 'Object-recognition-ratio-estimation module' produces the final output 'O.11'. A reference label 'P.1199(25)' is in the bottom right.
+
+**Figure 2 – Building blocks of the model**
+
+## 8.1 Model inputs
+
+The model receives the following input signal:
+
+### I.11: Codec-related factors
+
+Video resolution (pels); video bitrate (kbit/s); video frame rate (fps) per unit time for all modes.
+
+### I.12: Network-related factors
+
+- Video packet loss rate (%) per unit time for modes 0 and 2.
+- The number of lost video frames (frames) per unit time for Mode 1 and Mode 3. Since the video-frame-loss pattern (e.g., random, burst) is not considered, the value of this factor becomes the same when the number of lost frames in one unit of time is the same, even if the frame loss patterns are different, as shown in Figure 3.
+
+![Figure 3 shows three example loss patterns (Loss pattern 1, Loss pattern 2, and Loss pattern 3) over a sequence of video frames. Each pattern is defined by a 'One unit time' interval, indicated by a double-headed arrow above the frames. The frames are represented by boxes labeled 'Frame'. Lost frames are indicated by the word 'Lost' below the corresponding frame box. In Loss pattern 1, 5 frames are lost within the unit time. In Loss pattern 2, 6 frames are lost within the unit time. In Loss pattern 3, 5 frames are lost within the unit time. The text 'P.1199(25)' is visible in the bottom right corner of the diagram area.](d4af765160d04ecef538e5066006dc77_img.jpg)
+
+Loss pattern 1
+
+One unit time
+
+Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame
+
+Lost Lost Lost Lost Lost
+
+Loss pattern 2
+
+One unit time
+
+Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame
+
+Lost Lost Lost Lost Lost Lost
+
+Loss pattern 3
+
+One unit time
+
+Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame Frame
+
+Lost Lost Lost Lost Lost
+
+P.1199(25)
+
+Figure 3 shows three example loss patterns (Loss pattern 1, Loss pattern 2, and Loss pattern 3) over a sequence of video frames. Each pattern is defined by a 'One unit time' interval, indicated by a double-headed arrow above the frames. The frames are represented by boxes labeled 'Frame'. Lost frames are indicated by the word 'Lost' below the corresponding frame box. In Loss pattern 1, 5 frames are lost within the unit time. In Loss pattern 2, 6 frames are lost within the unit time. In Loss pattern 3, 5 frames are lost within the unit time. The text 'P.1199(25)' is visible in the bottom right corner of the diagram area.
+
+**Figure 3 – Example loss patterns with five lost video frames per unit**
+
+### I.13: Vehicle-related factors
+
+Autonomous vehicle velocity (km/h) per unit time. Note that this factor is used for Mode 2 and Mode 3.
+
+### I.GEN: A priori information for estimation module
+
+A priori information for the estimation module includes the following:
+
+- Display resolution
+- Packet loss error concealment
+- Codec
+  - Codec
+  - Profile
+  - GOP
+- Source-related factors
+  - Object colour
+  - Object size
+ Weather
+  - Timeframe
+- Network latency factors
+  - Glass-to-glass latency
+  - Control latency
+- Reaction time factors
+  - Required reaction time
+- Camera-related factors
+  - Camera mode
+  - Focal length
+  - Field of view
+  - Image stabilization
+  - Auto white balance delay
+
+- Lens flare
+- Road-related factors
+  - Road surface coefficient of friction (this is determined on the basis of weather).
+
+## 8.2 Model outputs
+
+### O.11: Output object-recognition ratio
+
+This model outputs the object-recognition ratio from 0 to 1. As described in clause 1, the object-recognition ratio is defined as the percentage of observers who can recognize objects within the time it is possible to prevent the vehicle from colliding with them when these objects appear a target distance away from the autonomous vehicle.
+
+# 9 The model algorithm
+
+The model in this Recommendation provides O.11 (object-recognition ratio) from 0 to 1. The model consists of four different modes, and the inputs differ between each mode. The following parameters are used in the description of the model:
+
+- $t$ : the unit time for video and network information acquisition
+- $b$ : the video bitrate per unit time.
+- $f$ : the video frame rate per unit time.
+- $r$ : the video resolution (the number of pels (*height* · *width*)) per unit time.
+- $v$ : the vehicle's velocity per unit time.
+- $l$ : the number of lost frames. It is directly measured in Mode 1 and Mode 3. By definition, since the number of lost frames is not measured in Mode 0 and Mode 2,  $l$  is estimated by packet loss rate.
+- $p$ : packet loss rate per unit time.
+- $X$ : object-recognition ratio if no packet loss or frame loss occurred.
+
+The object-recognition ratio O.11 is calculated as follows:
+
+$$O.11 = \frac{X}{1 + \left( -\frac{c_1}{f^{c_2}} \log \left( \frac{f \cdot t - l}{f \cdot t} \right) \right)^{c_3}}, \quad (1)$$
+
+$$X = A_1 - \frac{A_1}{1 + \left( \frac{b}{A_2} \right)^{A_3}}. \quad (2)$$
+
+In modes 0 and 2,  $l$  is not directly obtained, so it is calculated as follows:
+
+$$l = \left( 1 - (1 - p)^{\frac{c_{11} \cdot b}{f^{c_{12}}} + c_{13}} \right) f \cdot t, \quad (3)$$
+
+where  $c_1 - c_3$  and  $c_{11} - c_{13}$  are the coefficients used in the model.
+
+Equation (2) is common to all four modes. However, the definitions of  $A_1 - A_3$  differ depending on the modes. In Mode 0 and Mode 1, the vehicle's velocity cannot be used and  $A_1 - A_3$  are calculated without velocity. In Mode 2 and Mode 3, velocity can be used, and  $A_1 - A_3$  are calculated with velocity.
+
+In Mode 0 and Mode 1,  $A_1 - A_3$  are calculated as follows:
+
+$$A_1 = c_4 - \frac{c_5}{r}, \quad (4)$$
+
+$$A_2 = c_6 \cdot f^{c_7}, \quad (5)$$
+
+$$A_3 = c_8. \quad (6)$$
+
+In Mode 2 and Mode 3,  $A_1 - A_3$  are calculated as follows:
+
+$$A_1 = 1 - c_4 \cdot v^{c_5} - \frac{c_6}{r}, \quad (7)$$
+
+$$A_2 = (c_7 + c_8 \cdot v) \cdot f^{c_9}, \quad (8)$$
+
+$$A_3 = c_{10}, \quad (9)$$
+
+where  $c_4 - c_{10}$  are the coefficients used in the model.
+
+Equations used per mode are listed in Table 3.
+
+**Table 3 – Equations per mode**
+
+| Mode     | 0         | 1                      | 2                       | 3                      |
+|----------|-----------|------------------------|-------------------------|------------------------|
+| Equation | (1) – (6) | (1), (2) and (4) – (6) | (1) – (3) and (7) – (9) | (1), (2) and (7) – (9) |
+
+The sets of a priori information and coefficients are listed in Tables 4 to 7 for each mode.
+
+**Table 4 – A priori information and coefficients sets of Mode 0**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Coefficients         | $c_1$         | 10.16      | 14.56     | 8.375      |
+|                      | $c_2$         | 0.8337     | 1.021     | 0.2591     |
+|                      | $c_3$         | 2.198      | 4.060     | 1.000      |
+|                      | $c_4$         | 0.9576     | 0.8514    | 0.8282     |
+|                      | $c_5$         | 25.03      | 25.00     | 25.10      |
+|                      | $c_6$         | 43.25      | 44.18     | 15.29      |
+|                      | $c_7$         | 0.7807     | 0.8791    | 0.6083     |
+|                      | $c_8$         | 1.293      | 8.837     | 1.567      |
+|                      | $c_9$         | –          | –         | –          |
+|                      | $c_{10}$      | –          | –         | –          |
+|                      | $c_{11}$      | 0.03263    | 0.05923   | 0.01773    |
+|                      | $c_{12}$      | 0.8199     | 1.004     | 0.6581     |
+|                      | $c_{13}$      | 3.340      | 3.773     | 1.001      |
+
+**Table 5 – A priori information and coefficients sets of Mode 1**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Coefficients         | $c_1$         | 9.694      | 12.76     | 1.528      |
+|                      | $c_2$         | 0.9096     | 1.045     | 0.2608     |
+|                      | $c_3$         | 2.151      | 2.682     | 3.670      |
+|                      | $c_4$         | 0.9182     | 0.8658    | 0.8121     |
+|                      | $c_5$         | 25.13      | 25.00     | 0.000      |
+|                      | $c_6$         | 48.15      | 44.36     | 15.43      |
+|                      | $c_7$         | 0.7750     | 0.8795    | 0.5961     |
+|                      | $c_8$         | 1.749      | 8.335     | 1.692      |
+
+**Table 6 – A priori information and coefficients sets of Mode 2**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Coefficients         | $c_1$         | 8.581      | 12.67     | 9.727      |
+|                      | $c_2$         | 0.6270     | 0.9969    | 0.1048     |
+|                      | $c_3$         | 2.567      | 3.292     | 1.000      |
+|                      | $c_4$         | 0.00001010 | 0.0004655 | 0.0005926  |
+|                      | $c_5$         | 2.735      | 1.680     | 1.582      |
+|                      | $c_6$         | 7049       | 25.00     | 25.00      |
+|                      | $c_7$         | 23.12      | 37.34     | 0.000      |
+|                      | $c_8$         | 1.637      | 0.3471    | 2.185      |
+|                      | $c_9$         | 0.7658     | 0.8736    | 0.3047     |
+|                      | $c_{10}$      | 3.291      | 9.254     | 2.378      |
+|                      | $c_{11}$      | 0.01239    | 0.06560   | 0.01528    |
+|                      | $c_{12}$      | 0.6402     | 0.9800    | 0.7329     |
+|                      | $c_{13}$      | 2.670      | 3.641     | 0.6220     |
+
+**Table 7 – A priori information and coefficients sets of Mode 3**
+
+| A priori information | Timeframe     | Daytime             | Nighttime | Daytime    |
+|----------------------|---------------|---------------------|-----------|------------|
+|                      | Codec         | H.265/HEVC          |           | H.265/HEVC |
+|                      | GOP structure | All-intra           |           | IPP        |
+| Coefficients         | $c_1$         | 6.272               | 14.36     | 1.850      |
+|                      | $c_2$         | 0.7521              | 1.081     | 0.2903     |
+|                      | $c_3$         | 2.119               | 2.428     | 2.543      |
+|                      | $c_4$         | 0.00001710          | 0.0001527 | 0.0005926  |
+|                      | $c_5$         | 2.543               | 1.969     | 1.582      |
+|                      | $c_6$         | $1.009 \times 10^4$ | 25.00     | 0.0001008  |
+|                      | $c_7$         | 12.60               | 36.45     | 0.000      |
+|                      | $c_8$         | 1.960               | 0.3680    | 2.185      |
+|                      | $c_9$         | 0.7732              | 0.8776    | 0.3047     |
+|                      | $c_{10}$      | 3.062               | 8.797     | 2.378      |
+
+## Annex A
+
+### Calculation of target distance
+
+(This annex forms an integral part of this Recommendation.)
+
+In estimating the object-recognition ratio, several delays need to be considered. Figure A.1 shows the workflow of events from the observer at the monitoring centre recognizing objects in a surveillance video to the stopping of an autonomous vehicle.
+
+The surveillance video is encoded and sent to the monitoring centre via a radio access network. Therefore, there are delays associated with encoding and network latency. In addition, there is a decoding and display latency until the video is displayed on the display. The sum of these delays is called glass-to-glass latency ( $L_{G2G}$ ), which is defined in a priori information.
+
+Required reaction time ( $L_{reaction}$ ) is the time between an observer recognizing an object and sending a brake command to an autonomous vehicle, which is defined in a priori information. This time relates to the observer's ability, such as viewing angle or contrast sensitivity.  $L_{reaction}$  depends on how much time the designer of the autonomous driving system gives the observer to recognize the object. In general, an observer takes at least 0.5 seconds to react to an object, so this required reaction time must be considered at this point.
+
+Control latency ( $L_{control}$ ) is the time between an observer sending a brake command and the autonomous vehicle starting to stop, which is defined in a priori information. The sum of glass-to-glass latency required reaction time and control latency is defined as total latency ( $L_{total}$ ) in seconds, and this is given by the following equation.
+
+$$L_{total} = L_{G2G} + L_{reaction} + L_{control}$$
+
+The braking distance ( $D_{break}$ ) in metres is given in the following equation using the autonomous vehicle's velocity  $V$  km/h and coefficient of friction  $\mu$  ( $\mu$  is the parameter depending on the weather).
+
+$$D_{break} = \frac{V^2}{254\mu}$$
+
+Therefore, the target distance  $X$  in metres is given in the following equation.
+
+$$X = \frac{V \cdot L_{total}}{3.6} + D_{break}$$
+
+The object-recognition ratio is the percentage of observers who can recognize objects within the time it is possible to prevent the vehicle from colliding with them when these objects appear a target distance away from the autonomous vehicle.
+
+![A sequence diagram illustrating the workflow from an autonomous vehicle to a monitoring centre and back. It shows the flow of video data, the observer's recognition process, and the subsequent brake command. Latencies for network, decoder, reaction, and control are marked, along with the total latency and a critical range for object recognition.](76b0cd79baaedd942af4dc42f2e764b8_img.jpg)
+
+The diagram illustrates the workflow of events from the observer at the monitoring centre recognizing objects in a surveillance video to the stopping of an autonomous vehicle. The process involves the following components and steps:
+
+- Autonomous vehicle:** A red car icon on the left. It sends a surveillance video via a camera and radio access network. It also receives a brake command and eventually stops at a brake distance  $D_{brake}$ .
+- Monitoring centre:** Located below the vehicle's path, it contains a **Decoder** and an **Observer** (represented by a person icon at a computer). The video is decoded and displayed for the observer.
+- Observer:** Recognizes objects in the video and sends a **Send brake command** back to the vehicle.
+- Target distance:**  $X$  m, indicated by a long red double-headed arrow at the top between the vehicle and an obstacle (a person walking away from a cube).
+- Latencies:**
+  - Network and encoder latency:** The time taken for the video to reach the decoder.
+  - Decoder and display latency:** The time taken to decode and display the video.
+  - Glass-to-glass latency:  $L_{GG}$ :** The time from video capture to display.
+  - Required reaction time:  $L_{reaction}$ :** The time the observer needs to recognize objects and decide to brake.
+  - Control latency:  $L_{control}$ :** The time taken for the brake command to reach the vehicle.
+  - Total latency:  $L_{total}$ :** The sum of all latencies from capture to vehicle stop.
+- Critical Range:** A red double-headed arrow labeled "Within this range, the observer must be able to recognize objects" spans from the start of the network/encoder latency to the start of the control latency.
+- Final State:** The autonomous vehicle stops at a **Brake distance:  $D_{brake}$**  from the obstacle.
+
+P.1199(25)
+
+A sequence diagram illustrating the workflow from an autonomous vehicle to a monitoring centre and back. It shows the flow of video data, the observer's recognition process, and the subsequent brake command. Latencies for network, decoder, reaction, and control are marked, along with the total latency and a critical range for object recognition.
+
+**Figure A.1 – The workflow of events from the observer at the monitoring centre recognizing objects in a surveillance video to the stopping of an autonomous vehicle**
+
+## Appendix I
+
+### Performance figures
+
+(This appendix does not form an integral part of this Recommendation.)
+
+In this appendix, the root mean squared error (RMSE) and Pearson correlation coefficient (PCC) are reported in Tables I.1 to I.4. The parametric object-recognition-estimation model per mode was validated on 296 daytime processed video sequences (PVSs) and 296 nighttime PVSs.
+
+The dataset of the object-recognition ratio is obtained by the subjective assessment test for the recognition task. In the subjective assessment, the participants watch surveillance videos in which objects appear at the target distance, and it is recorded whether the participants can recognize objects within the required reaction time. The object-recognition ratio of the dataset is the percentage of participants who can recognize objects within the time. The subjective assessment methods described in [b-ITU-T P.912] can be applied to the timed task method.
+
+**Table I.1 – RMSE and PCC of the model in Mode 0**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Score                | RMSE          | 0.254      | 0.216     | 0.341      |
+|                      | PCC           | 0.547      | 0.687     | 0.434      |
+
+**Table I.2 – RMSE and PCC of the model in Mode 1**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Score                | RMSE          | 0.247      | 0.202     | 0.292      |
+|                      | PCC           | 0.583      | 0.741     | 0.636      |
+
+**Table I.3 – RMSE and PCC of the model in Mode 2**
+
+| A priori information | Timeframe     | Daytime    | Nighttime | Daytime    |
+|----------------------|---------------|------------|-----------|------------|
+|                      | Codec         | H.265/HEVC |           | H.265/HEVC |
+|                      | GOP structure | All-intra  |           | IPP        |
+| Score                | RMSE          | 0.212      | 0.176     | 0.295      |
+|                      | PCC           | 0.715      | 0.819     | 0.633      |
+
+**Table I.4 – RMSE and PCC of the model in Mode 3**
+
+| <b>A priori<br/>information</b> | <b>Timeframe</b>     | <b>Daytime</b>    | <b>Nighttime</b> | <b>Daytime</b>    |
+|---------------------------------|----------------------|-------------------|------------------|-------------------|
+|                                 | <b>Codec</b>         | <b>H.265/HEVC</b> |                  | <b>H.265/HEVC</b> |
+|                                 | <b>GOP structure</b> | <b>All-intra</b>  |                  | <b>IPP</b>        |
+| <b>Score</b>                    | <b>RMSE</b>          | 0.204             | 0.164            | 0.231             |
+|                                 | <b>PCC</b>           | 0.743             | 0.861            | 0.795             |
+
+## Appendix II
+
+### Camera-related factors
+
+(This appendix does not form an integral part of this Recommendation.)
+
+This appendix presents the camera-related factors of the model used in the validation experiment. These values of camera-related factors of each daytime and nighttime PVSs in the validation experiment are shown in Table II.1. In the validation experiment, these values are used as representative values of taking surveillance videos for autonomous driving so that objects can be recognized in the surveillance videos with an appropriate size and appearance.
+
+**Table II.1 – Parameters of camera-related factors of each daytime and nighttime PVSs**
+
+| Timeframe                       | Daytime              | Nighttime |
+|---------------------------------|----------------------|-----------|
+| Camera mode                     | Fisheye disable mode |           |
+| Camera focal length             | 24 mm                | 14mm      |
+| Camera field of view (diagonal) | 95.5°                | 114°      |
+| Camera image stabilization      | Yes                  |           |
+| Camera auto white balance delay | No                   |           |
+| Lens flare                      | No                   |           |
+
+## Appendix III
+
+### Guidelines for the use case of the object-recognition ratio
+
+(This appendix does not form an integral part of this Recommendation.)
+
+This appendix provides two use cases demonstrating the application of the object-recognition ratio in real-world autonomous driving services. The first is for observers at a monitoring centre to avoid continued monitoring when the quality of surveillance video is poor. Specifically, the object-recognition ratio obtained from this Recommendation can support observers at the monitoring centre by triggering alerts when the object-recognition ratio falls below a predefined threshold. The second aims to extract areas within the autonomous driving area where network bandwidth is reduced in terms of the object-recognition ratio. If there are areas within the autonomous driving area where the object-recognition ratio is continuously insufficient, safe autonomous driving cannot be achieved. By measuring the object-recognition ratio within the autonomous driving area, network operators can detect the areas where the wireless network is poor for autonomous driving.
+
+## III.1 Use Case 1: Alerting observers at the monitoring centre
+
+In autonomous driving systems, an autonomous vehicle is equipped with object-detection technology to detect objects around the autonomous vehicle. On the other hand, current object-detection technology is challenging to use on local streets because various objects appear, such as traffic signs or pedestrians. Therefore, autonomous driving safety needs to be ensured in real time by observers at the monitoring centre. If an observer recognizes objects that interfere with driving, they need to react and brake the autonomous vehicle immediately when an object appears. However, since the surveillance video is encoded and delivered through a network, video quality may degrade due to limited wireless bandwidth, making it difficult for observers to recognize objects in the video. The observer has difficulty continuously judging whether the object can be recognized from surveillance video in real time. This clause describes a use case in which an alert is triggered when object recognition becomes difficult due to degraded video quality.
+
+#### III.1.1 Operational flow
+
+The surveillance system and how alerts are triggered for observers at the monitoring centre are shown in Figure III.1. The detailed flow of Use Case 1 is described in clauses III.1.1.1 to III.1.1.5.
+
+![Figure III.1: An overview of the surveillance system and the operational flow of triggering alerts to observers at the monitoring centre. The diagram shows a red autonomous vehicle on a road with pedestrians. A surveillance camera and encoder on the vehicle send video packets and vehicle velocity information via a wireless network to a monitoring centre. Inside the monitoring centre, a decoder receives the packets and displays the surveillance video on a monitor for an observer. Simultaneously, input parameters are extracted and sent to an object-recognition-ratio-calculation device. This device calculates the object-recognition ratio per unit time using the input parameters and the vehicle's velocity. The calculated ratio is displayed on another monitor, which also shows a graph of the ratio over time. If the ratio falls below a predefined threshold (e.g., 80%), an alert is triggered, indicated by a flashing light and a message on the monitor.](c914f51f4427bc672dd0526cfc90ebe9_img.jpg)
+
+The diagram illustrates the flow of data in a surveillance system. On the left, an autonomous vehicle is shown on a road with pedestrians. A surveillance camera and encoder are mounted on the vehicle. The vehicle sends video packets and its velocity (Mode 2 and 3) through a wireless network to a monitoring centre. Inside the monitoring centre, the video packets are received by a decoder, which then displays the surveillance video on a monitor for an observer. The decoder also extracts input parameters, which are sent to an object-recognition-ratio-calculation device. This device calculates the object-recognition ratio per unit time, using the input parameters and the vehicle's velocity. The calculated ratio is displayed on a second monitor, which also shows a graph of the ratio over time. A threshold (e.g., 80%) is indicated on this monitor. If the ratio falls below the threshold, an alert is triggered, indicated by a flashing light and a message on the monitor.
+
+Figure III.1: An overview of the surveillance system and the operational flow of triggering alerts to observers at the monitoring centre. The diagram shows a red autonomous vehicle on a road with pedestrians. A surveillance camera and encoder on the vehicle send video packets and vehicle velocity information via a wireless network to a monitoring centre. Inside the monitoring centre, a decoder receives the packets and displays the surveillance video on a monitor for an observer. Simultaneously, input parameters are extracted and sent to an object-recognition-ratio-calculation device. This device calculates the object-recognition ratio per unit time using the input parameters and the vehicle's velocity. The calculated ratio is displayed on another monitor, which also shows a graph of the ratio over time. If the ratio falls below a predefined threshold (e.g., 80%), an alert is triggered, indicated by a flashing light and a message on the monitor.
+
+**Figure III.1 – An overview of the surveillance system and the operational flow of triggering alerts to observers at the monitoring centre**
+
+##### III.1.1.1 Sending video and velocity information
+
+The surveillance camera mounted on the autonomous vehicle captures surveillance video during autonomous driving. The real-time encoder in the autonomous vehicle encodes the surveillance video. The encoded video is transmitted via a wireless network by sending video packets to the decoder at the monitoring centre. If the vehicle's velocity is available, it is also sent to the object-recognition-ratio-calculation device at the monitoring centre, and the velocity can be used to calculate the object-recognition ratio.
+
+##### III.1.1.2 Displaying surveillance video
+
+The decoder receives video packets transmitted from the autonomous vehicle. These packets are decoded into surveillance video frames, which are then displayed in real time on the surveillance video monitor for observers. This enables observers to monitor the driving environment visually.
+
+##### III.1.1.3 Extract input parameters
+
+The decoder at the monitoring centre extracts input parameters from the received video stream, including video bitrate, framerate, resolution, packet loss information and frozen video frame information. These extracted input parameters are then transmitted to the object-recognition-ratio-calculation device.
+
+##### III.1.1.4 Calculating the object-recognition ratio
+
+The object-recognition-ratio-calculation device calculates the object-recognition ratio per unit time using input parameters. In Modes 2 and 3, the vehicle's velocity is also used to calculate the object-recognition ratio. The calculated object-recognition ratio is displayed on the object-recognition-ratio monitor at the monitoring centre.
+
+##### III.1.1.5 Threshold-based alerting
+
+The calculated object-recognition ratio is chronologically displayed on the monitoring system interface. If the object-recognition ratio falls below the predefined threshold, which is determined on
+
+the basis of the requirements of the monitoring system provider, an alert is triggered to notify the observer of a safety risk due to video quality degradation. The observer should take appropriate action depending on the situation, such as reducing the vehicle's speed or initiating a controlled stop of the vehicle to ensure safety.
+
+## III.2 Use Case 2: Detecting object-recognition-ratio-degradation areas
+
+The wireless network bandwidth fluctuates due to environmental factors such as surrounding buildings or the time of day, and these fluctuations of bandwidth may lead to increased risks in specific areas or during certain times for the remote monitoring. If surveillance video quality degrades due to degraded wireless bandwidth and the object-recognition ratio falls, an observer may not be able to recognize objects that interfere with driving. Therefore, stable object-recognition performance is essential for safe autonomous driving.
+
+In this use case, object-recognition ratio data is used to estimate and visualize spatial and temporal variations in recognition performance. This allows network operators to understand trends in object-recognition performance across autonomous driving areas.
+
+#### III.2.1 Operational flow
+
+An example of the detection of object-recognition-ratio-degradation area using the object-recognition ratio is shown in Figure III.2, and the detailed flow is described in the following clause.
+
+![Figure III.2: An example of the detection of object-recognition-ratio-degradation area. The diagram is divided into two main sections. Section 1, 'Object-recognition-ratio-data collection in autonomous driving area', shows a map with a green rectangular boundary and a bus icon. Section 2, 'Detection of object-recognition-ratio-degradation area', contains two sub-panels: 'Daytime' and 'Night-time'. Each sub-panel shows a map with red heat spots indicating degradation areas. A label 'The object-recognition ratio degradation area' points to these spots. A small text 'P.1199(25)' is in the bottom right corner.](171115f072e42b379238ed0dd438e9d7_img.jpg)
+
+1. Object-recognition-ratio-data collection in autonomous driving area
+
+2. Detection of object-recognition-ratio-degradation area
+
+Daytime
+
+Night-time
+
+The object-recognition ratio degradation area
+
+P.1199(25)
+
+Figure III.2: An example of the detection of object-recognition-ratio-degradation area. The diagram is divided into two main sections. Section 1, 'Object-recognition-ratio-data collection in autonomous driving area', shows a map with a green rectangular boundary and a bus icon. Section 2, 'Detection of object-recognition-ratio-degradation area', contains two sub-panels: 'Daytime' and 'Night-time'. Each sub-panel shows a map with red heat spots indicating degradation areas. A label 'The object-recognition ratio degradation area' points to these spots. A small text 'P.1199(25)' is in the bottom right corner.
+
+Figure III.2 – An example of the detection of object-recognition-ratio-degradation area
+
+##### III.2.1.1 Object-recognition-ratio-data collection
+
+To identify areas with degraded object-recognition ratios within the autonomous driving area, the system provider test-drives vehicles equipped with surveillance cameras. As the vehicle moves through the area, surveillance video is captured along with the corresponding location and timestamp data. The set of location information where the vehicle passes through, timestamps and the calculated object-recognition-ratio data is stored for detecting the object-recognition-ratio-degradation areas.
+
+##### III.2.1.2 Detection of object-recognition-ratio-degradation areas
+
+Using the collected dataset of object-recognition ratios, the network operator can detect areas or times where object-recognition performance is degraded. This enables the identification of specific areas and times where recognition performance falls below acceptable levels. While the specific methods for improving degraded areas are beyond the scope of this document, network operators may take actions such as installing additional base stations or adjusting tilt angles to enhance coverage.
+
+# Bibliography
+
+- [b-ITU-T P.912] Recommendation ITU-T P.912 (2016), *Subjective video quality assessment methods for recognition tasks*.
+- [b-Cai] Cai, Y., Wang, J., Chen, H., et al. (2021), *YOLOv4-5D: An Effective and Efficient Object Detector for Autonomous Driving*, IEEE Transactions on Instrumentation and Measurement, Vol. 70, pp. 1–13.
+- [b-French Gov.] French Government (2021), *The French strategy for the development of automated road mobility 2020-2022*, Informal Document GRVA-099-03.
+- [b-German Gov.] Bundesministerium für Verkehr und digitale Infrastruktur (2021), *Entwurf eines Gesetzes zur Änderung des Straßenverkehrsgesetzes und des Pflichtversicherungsgesetzes – Gesetz zum autonomen Fahren*.
+- [b-Ikeuchi] Ikeuchi, H. (2022), *Initiatives for Realization of Automated Driving by Japan Police*, ITS World Congress.
+
+
+
+
+
+## SERIES OF ITU-T RECOMMENDATIONS
+
+|                 |                                                                                                                                                           |
+|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------|
+| Series A        | Organization of the work of ITU-T                                                                                                                         |
+| Series D        | Tariff and accounting principles and international telecommunication/ICT economic and policy issues                                                       |
+| Series E        | Overall network operation, telephone service, service operation and human factors                                                                         |
+| Series F        | Non-telephone telecommunication services                                                                                                                  |
+| Series G        | Transmission systems and media, digital systems and networks                                                                                              |
+| Series H        | Audiovisual and multimedia systems                                                                                                                        |
+| Series I        | Integrated services digital network                                                                                                                       |
+| Series J        | Cable networks and transmission of television, sound programme and other multimedia signals                                                               |
+| Series K        | Protection against interference                                                                                                                           |
+| Series L        | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant |
+| Series M        | Telecommunication management, including TMN and network maintenance                                                                                       |
+| Series N        | Maintenance: international sound programme and television transmission circuits                                                                           |
+| Series O        | Specifications of measuring equipment                                                                                                                     |
+| <b>Series P</b> | <b>Telephone transmission quality, telephone installations, local line networks</b>                                                                       |
+| Series Q        | Switching and signalling, and associated measurements and tests                                                                                           |
+| Series R        | Telegraph transmission                                                                                                                                    |
+| Series S        | Telegraph services terminal equipment                                                                                                                     |
+| Series T        | Terminals for telematic services                                                                                                                          |
+| Series U        | Telegraph switching                                                                                                                                       |
+| Series V        | Data communication over the telephone network                                                                                                             |
+| Series X        | Data networks, open system communications and security                                                                                                    |
+| Series Y        | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities                               |
+| Series Z        | Languages and general software aspects for telecommunication systems                                                                                      |

marked/P/T-REC-P.1202-201210-I_PDF-E/raw.md

@@ -0,0 +1,595 @@
+
+
+**ITU-T**
+
+TELECOMMUNICATION  
+STANDARDIZATION SECTOR  
+OF ITU
+
+**P.1202**
+
+(10/2012)
+
+SERIES P: TERMINALS AND SUBJECTIVE AND  
+OBJECTIVE ASSESSMENT METHODS
+
+Models and tools for quality assessment of streamed  
+media
+
+---
+
+**Parametric non-intrusive bitstream assessment  
+of video media streaming quality**
+
+Recommendation ITU-T P.1202
+
+# ITU-T P-SERIES RECOMMENDATIONS **TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS**
+
+|                                                                                               |               |               |
+|-----------------------------------------------------------------------------------------------|---------------|---------------|
+| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series        | P.10          |
+| Voice terminal characteristics                                                                | Series        | P.30          |
+|                                                                                               |               | P.300         |
+| Reference systems                                                                             | Series        | P.40          |
+| Objective measuring apparatus                                                                 | Series        | P.50          |
+|                                                                                               |               | P.500         |
+| Objective electro-acoustical measurements                                                     | Series        | P.60          |
+| Measurements related to speech loudness                                                       | Series        | P.70          |
+| Methods for objective and subjective assessment of speech quality                             | Series        | P.80          |
+|                                                                                               |               | P.800         |
+| Audiovisual quality in multimedia services                                                    | Series        | P.900         |
+| Transmission performance and QoS aspects of IP end-points                                     | Series        | P.1000        |
+| Communications involving vehicles                                                             | Series        | P.1100        |
+| <b>Models and tools for quality assessment of streamed media</b>                              | <b>Series</b> | <b>P.1200</b> |
+| Telemeeting assessment                                                                        | Series        | P.1300        |
+| Statistical analysis, evaluation and reporting guidelines of quality measurements             | Series        | P.1400        |
+
+*For further details, please refer to the list of ITU-T Recommendations.*
+
+## Recommendation ITU-T P.1202
+
+### Parametric non-intrusive bitstream assessment of video media streaming quality
+
+## Summary
+
+Recommendation ITU-T P.1202 provides an overview of algorithmic models for non-intrusive monitoring of the video quality of IP-based video services based on packet-header and bitstream information. The ITU-T P.1202-series of Recommendations addresses two application areas:
+
+- ITU-T P.1202.1 specifies the model algorithm for the lower resolution (LR) application area, including services such as mobile TV.
+- ITU-T P.1202.2 specifies the model algorithm for the higher resolution (HR) application area, which includes services such as IPTV.
+
+The ITU-T P.1202 model algorithms are no-reference (i.e., non-intrusive) models which operate by analysing packet header and bitstream information as available from respective packet trace data provided to the model algorithms in the packet capture format (PCAP). Further input information on more general aspects of the stream, which may not be available from packet header and bitstream information, is provided to the model algorithm out-of-band, for example in the form of stream-specific side information.
+
+ITU-T P.1202.1 describes one model, the model for the LR application area. ITU-T P.1202.2 describes two models for the HR application area corresponding to two modes: mode 1 and mode 2, which are both no-reference (i.e., non-intrusive) models. Mode 1 refers to a parsing mode; the model operates by analysing information in the video bitstream without fully decoding the bitstream (i.e., no pixel information is used) for MOS estimation. Mode 2 refers to a full decoding mode, in addition to the bitstream information which mode 1 uses, the model can also decode parts or all of the video bitstream (i.e., pixel information is used) for MOS estimation. Further client specific information, such as concealment type, is provided to the algorithm out-of-band, for example in the form of stream-specific side information. As output, the model algorithms provide individual estimates of video quality in terms of the five-point absolute category rating (ACR) mean opinion score (MOS). Further, diagnostic information on causes of quality degradations can be made available, too, since different types of performance parameters are derived during model calculations.
+
+Complementary to the ITU-T P.1202 models, there are two further models described in Recommendations ITU-T P.1201.1 and ITU-T P.1201.2. The respective entry-Recommendation for these models is ITU-T P.1201. It describes packet-header-only-based video, and audio and audiovisual quality models. The main differences with ITU-T P.1202 can be summarized as follows:
+
+- The ITU-T P.1201 models provide audio, video and audiovisual quality estimates, while the ITU-T P.1202-only models provide video quality estimates.
+
+The ITU-T P.1201 models use packet header information, while the ITU-T P.1202 models exploit further bitstream information, such as coding-related information. As a consequence, the ITU-T P.1202 models can be more accurate in their quality predictions. In turn, they require non-encrypted streams to enable access to payload information. Since the ITU-T P.1202 models are more complex, they also require more computational power to estimate the video quality.
+
+## History
+
+| Edition | Recommendation             | Approval   | Study Group |
+|---------|----------------------------|------------|-------------|
+| 1.0     | ITU-T P.1202               | 2012-10-14 | 12          |
+| 1.1     | ITU-T P.1202 (2012) Amd. 1 | 2013-03-28 | 12          |
+
+## Keywords
+
+Audio, audiovisual, IPTV, mean opinion score (MOS), mobile TV, monitoring, multimedia, QoE, video.
+
+## FOREWORD
+
+The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis.
+
+The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics.
+
+The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1.
+
+In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC.
+
+## NOTE
+
+In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency.
+
+Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party.
+
+## INTELLECTUAL PROPERTY RIGHTS
+
+ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process.
+
+As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at <http://www.itu.int/ITU-T/ipr/>.
+
+© ITU 2013
+
+All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU.
+
+## Table of Contents
+
+|                                                                                           | Page |
+|-------------------------------------------------------------------------------------------|------|
+| 1 Scope .....                                                                             | 1    |
+| 2 References.....                                                                         | 4    |
+| 3 Definitions .....                                                                       | 4    |
+| 3.1 Terms defined elsewhere .....                                                         | 4    |
+| 3.2 Terms defined in this Recommendation.....                                             | 4    |
+| 4 Abbreviations and acronyms .....                                                        | 5    |
+| 5 Conventions .....                                                                       | 5    |
+| 6 Areas of application.....                                                               | 6    |
+| 6.1 Application range for the models .....                                                | 6    |
+| 6.2 Modes of operation.....                                                               | 8    |
+| 7 Model input interfaces .....                                                            | 10   |
+| 8 Model output information and performance details .....                                  | 14   |
+| 9 Description of the ITU-T P.1202 model algorithm .....                                   | 15   |
+| Appendix I – Detailed performance figures for the ITU-T P.1202.1 algorithm.....           | 16   |
+| Appendix II – Detailed performance figures for the ITU-T P.1202.2 mode 1 algorithm.....   | 17   |
+| Appendix III – Detailed performance figures for the ITU-T P.1202.2 mode 2 algorithm ..... | 18   |
+
+
+
+## Parametric non-intrusive bitstream assessment of video media streaming quality
+
+# 1 Scope
+
+This Recommendation describes recommended objective models for non-intrusive monitoring of the video quality of IP-based video services based on packet-header and bitstream information. This Recommendation addresses two application areas:
+
+- [ITU-T P.1202.1] specifies the model algorithm for the lower resolution (LR) application area, including services such as mobile TV.
+- [ITU-T P.1202.2] specifies the model algorithm for the higher resolution (HR) application area, which includes services such as IPTV. This application area is currently under study but not yet completed.
+
+These models are restricted to information contained in packet headers, the packet bitstream (payload information), prior and static knowledge about the media stream and dynamic buffering information from the client.
+
+[ITU-T P.1202.1] consists of one model. [ITU-T P.1202.2] consists of two models corresponding to two modes: mode 1 and mode 2, which both are no-reference (i.e., non-intrusive) models. Mode 1 refers to a parsing mode; the model operates by analysing information in the video bitstream without fully decoding the bitstream (i.e., no pixel information is used) for MOS estimation. Mode 2 refers to a full decoding mode, in addition to the bitstream information which mode 1 uses, the model can also decode parts or all of the video bitstream (i.e., pixel information is used) for mean opinion score (MOS) estimation. Further client specific information, such as concealment type, is provided to the algorithm out-of-band, for example in the form of stream specific side information.
+
+These models predict video quality in terms of MOSs on a five-point ACR scale (see [ITU-T P.910]).
+
+The primary application for these models is the monitoring of transmission quality during service operation or for maintenance purposes. The ITU-T P.1202 model may be deployed both in end-point locations and at mid-network monitoring points. The location of the model and the location of the measurement probe together determine the *mode of operation*, as described in more detail in clause 6.1.
+
+The primary quality prediction made by such models is based on the payload of the stream being analysed. Therefore, this Recommendation can provide a comprehensive evaluation of quality as perceived by a particular end-user because its scores can reflect the impairments on the coding and the Internet Protocol (IP) network being measured, which differ from user to user. This Recommendation cannot provide a comprehensive evaluation of video quality as perceived by a particular end-user, because its scores reflect the impairments due to encoding and the subsequent IP network being assessed, which may only be one part of the end-to-end connection. An explicit inclusion of processing steps such as content contribution from e.g., satellite networks, display properties etc. are not considered. See Table 3 for more information. Further, the quality-impact due to a specific video encoder implementation or a specific decoder-side packet loss concealment implementation is not explicitly addressed. Instead, the models have been developed for a set of dedicated service implementations, which are assumed to be meaningful representations of today's IP-based streaming video services. As a consequence, however, in case of significant deviations of a given service being assessed from the service configurations used for developing this standard, it is possible to obtain high quality scores with this Recommendation and yet to have a poor quality of the stream as it is perceived by actual users.
+
+As a consequence, this Recommendation can be used for applications such as:
+
+- in-service quality monitoring for specific IP-based audiovisual services, as specified in more detail in Tables 4 to 7;
+- benchmarking of different service implementations. However, it cannot be used for direct benchmarking of different decoder implementations. The implementations that can be assessed with ITU-T P.1202 include the encoding aspects and potential packet loss.
+
+The application areas of the ITU-T P.1202 model algorithms are summarized in Tables 1, 2, and 3 below:
+
+**Table 1 – Application areas, test factors, and coding technologies for which [ITU-T P.1202.1] has been verified and is known to produce reliable results.  
+For details about the settings, see clause 6**
+
+| <b>ITU-T P.1202.1 lower resolution (LR)</b>                                                                       | <b>Higher resolution (HR)</b> |
+|-------------------------------------------------------------------------------------------------------------------|-------------------------------|
+| <b>Applications for which the models are intended</b>                                                             |                               |
+| In-service monitoring of video UDP-based streaming                                                                | Under study                   |
+| Performance and quality assessment of live networks including the effect due to encoding and transmission errors  | Under study                   |
+| <b>Test factors for which the models have been validated</b>                                                      |                               |
+| Encoding (compression) degradation of video with a variety of bitrates<br>Video: 50 – 6000 kbit/s                 | Under study                   |
+| Packet loss degradation of video (both random and bursty packet loss patterns)                                    | Under study                   |
+| Re-buffering degradation                                                                                          | –                             |
+| Video contents of different spatio-temporal complexity                                                            | Under study                   |
+| Different video keyframe and frame-rates<br>Frame rates: 12.5-30 Hz<br>GOP lengths (1/keyframe rate): 2-10 s      | Under study                   |
+| Different video resolutions: HVGA, QVGA, QCIF                                                                     | Under study                   |
+| Different decoder-side packet loss concealment strategies (freezing with skipping, one/multiple slices per frame) | Under study                   |
+| <b>Coding technologies on which the models have been trained</b>                                                  |                               |
+| Video: ITU-T H.264/AVC baseline profile                                                                           | Under study                   |
+
+**Table 2 – Application areas, test factors, and coding technologies for which further investigation of ITU-T P.1202 models is needed**
+
+| <b>ITU-T P.1202.1 lower resolution (LR)</b>                                                                                                  | <b>Higher resolution (HR)</b> |
+|----------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------|
+| <b>Applications for which the models can be used, but the results may not be reliable</b>                                                    |                               |
+| In-service monitoring of live network video TCP based streaming (assuming that parameter extraction from TCP based streaming is implemented) | Under study                   |
+| Direct comparison/benchmarking of encoder implementations, and thus of services that employ different encoder implementations                | Under study                   |
+| <b>Test factors for which the models can be used but the results may not be reliable</b>                                                     |                               |
+| –                                                                                                                                            | –                             |
+| <b>Coding technologies for which the models can be used but the results may not be reliable</b>                                              |                               |
+| –                                                                                                                                            | –                             |
+
+**Table 3 – Application areas, test factors, and coding technologies for which ITU-T P.1202 models are not intended to be used**
+
+| <b>ITU-T P.1202.1 lower resolution (LR)</b>                                                                                   | <b>Higher resolution (HR)</b> |
+|-------------------------------------------------------------------------------------------------------------------------------|-------------------------------|
+| <b>Applications for which the models are not intended</b>                                                                     |                               |
+| Direct comparison/benchmarking of decoder implementations, and thus of services that employ different decoder implementations | Under study                   |
+| Evaluation of visual quality including display/device properties                                                              | Under study                   |
+| <b>Test factors for which the models are not intended</b>                                                                     |                               |
+| Video streaming with significant rate adaptation (such as used in dynamic adaptive streaming over HTTP (DASH))                | Under study                   |
+| Transcoding situations                                                                                                        | Under study                   |
+| The effects of noise, delay, colour correctness                                                                               | Under study                   |
+| Audio visual streaming                                                                                                        | Under study                   |
+| <b>Coding technologies for which the models are not intended</b>                                                              |                               |
+| [ITU-T H.261], MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.265, etc.                                                                 |                               |
+
+# 2 References
+
+The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation.
+
+- [ITU-T H.264] Recommendation ITU-T H.264 (2011), *Advanced video coding for generic audiovisual services*.
+- [ITU-T P.800.1] Recommendation ITU-T P.800.1 (2006), *Mean Opinion Score (MOS) terminology*.
+- [ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*.
+- [ITU-T P.1201] Recommendation ITU-T P.1201 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality*.
+- [ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – lower resolution application area*.
+- [ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – higher resolution application area*.
+- [ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – lower resolution application area*.
+- [ITU-T P.1202.2] Recommendation ITU-T P.1202.2 (2013), *Parametric non-intrusive bitstream assessment of video media streaming quality – Higher resolution application area*.
+- [ITU-T P.1401] Recommendation ITU-T P.1401 (2012), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*.
+
+# 3 Definitions
+
+## 3.1 Terms defined elsewhere
+
+This Recommendation uses the following term defined elsewhere:
+
+- 3.1.1 **mean opinion score (MOS)**: [ITU-T P.800.1].
+
+## 3.2 Terms defined in this Recommendation
+
+This Recommendation defines the following terms:
+
+- 3.2.1 **model, model algorithm**: An algorithm with the purpose of estimating the subjective (perceived) quality of a media sequence.
+- 3.2.2 **sequence**: A short decoded audio, video or audiovisual portion of a stream, typically shorter than 30 seconds.
+- 3.2.3 **bitstream**: The part of an IP-based transmission where the actual audiovisual, video or audio content is available in encoded and packetized form.
+
+**3.2.4 compression artefacts:** Artefacts introduced due to lossy compression of the encoding process.
+
+**3.2.5 slicing artefacts:** Artefacts introduced when packet losses are concealed using a packet-loss concealment (PLC) scheme trying to repair erroneous frames.
+
+**3.2.6 freezing artefacts:** Artefacts introduced when the packet-loss concealment (PLC) scheme of the receiver replaces the erroneous frames (either due to packet loss or error propagation) with the previous error free frame until a decoded picture without errors has been received. Since the erroneous frames are not displayed, this type of artefact is also referred to as freezing with skipping.
+
+**3.2.7 rebuffering artefacts:** Artefacts coming from rebuffering events at the client side, which could be a result of video data arriving late. Usually, rebuffering events are indicated to the viewer, e.g., in the form of a spinning wheel. This is also referred to as freezing without skipping.
+
+# 4 Abbreviations and acronyms
+
+This Recommendation uses the following abbreviations and acronyms:
+
+|       |                                        |
+|-------|----------------------------------------|
+| DASH  | Dynamic Adaptive Streaming over HTTP   |
+| GOP   | Group of Pictures                      |
+| HD    | High Definition (television)           |
+| HRC   | Hypothetical Reference Circuit         |
+| HVGA  | Half Video Graphics Array              |
+| IP    | Internet Protocol                      |
+| MBAFF | Macroblock-Adaptive Frame-Field        |
+| MBMS  | Multimedia Broadcast Multicast Service |
+| MOS   | Mean Opinion Score                     |
+| MPEG  | Motion Pictures Expert Group           |
+| NTSC  | National Television Standard Committee |
+| PAL   | Phase Alternating Line                 |
+| PCAP  | Packet Capture format                  |
+| PSS   | Packet Switched Streaming              |
+| PVS   | Processed Video Sequence               |
+| QCIF  | Quarter Common Intermediate Format     |
+| QoE   | Quality of Experience                  |
+| QVGA  | Quarter Video                          |
+| RTP   | Real-time Transport Protocol           |
+| SD    | Standard Definition                    |
+| SRC   | Source Reference Channel or Circuit    |
+| UDP   | User Datagram Protocol                 |
+
+# 5 Conventions
+
+None.
+
+# 6 Areas of application
+
+The two application areas for ITU-T P.1202 are:
+
+- [ITU-T P.1202.1] (lower resolution mode (LR)):  
+QCIF-QVGA-HVGA, mostly for mobile TV and streaming with the sub-application areas:
+  - Linear mobile TV over RTP (includes mobile TV over a 3G mobile network with MBMS and with unicast, transport over RTP/UDP/IP).
+  - Multimedia streaming (includes 3GPP PSS with transport over RTP/UDP/IP).
+- Higher resolution mode, (HR): SD and HD television, mostly for IPTV with the sub-application areas (this mode is currently under study):
+  - Linear broadcast TV (includes transmission over MPEG2-TS/RTP/UDP/IP, and is assumed to be applicable to MPEG2-TS/UDP/IP and RTP/UDP/IP transport with similar, but so far unverified accuracy as compared to MPEG2-TS/RTP/UDP/IP).
+  - Video on-demand (includes transmission over MPEG2-TS/RTP/UDP/IP, and is assumed to be applicable to MPEG2-TS/UDP/IP and RTP/UDP/IP transport with similar, but so far unverified accuracy as compared to MPEG2-TS/RTP/UDP/IP).
+
+## 6.1 Application range for the models
+
+Table 4 below shows the application range of the models based on what the models have actually been trained for. Note that all cases represent the CC mode of operation, see clause 6.2 for more details about the modes.
+
+**Table 4 – Factors and application ranges of the ITU-T P.1202 model algorithms**
+
+|                                | <b>ITU-T P.1202.1 lower resolution (LR)</b>                                                                                                                                                                                                                                                                                                                                       | <b>Higher resolution (HR)</b> |
+|--------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------|
+| <b>Application information</b> | <b>Value range, unit</b>                                                                                                                                                                                                                                                                                                                                                          |                               |
+| Sequence duration (Ts)         | The model has been validated on source sequence lengths of:<br>10 s: no rebuffering<br>16 s: rebuffering<br>No rebuffering: PVS length = SRC length<br>Rebuffering: PVS length = SRC length + rebuffering length<br>(no rebuffering at end and start)<br>It is expected that the model will give reliable prediction results for sequence durations within the range 8-24 seconds | Under study                   |
+| Packetization                  | 3GPP MBMS, PSS or using RTSP directly (all three over RTP/UDP/IP)                                                                                                                                                                                                                                                                                                                 | Under study                   |
+| Video codec                    | ITU-T H.264/AVC baseline profile                                                                                                                                                                                                                                                                                                                                                  | Under study                   |
+| Video size                     | QCIF, QVGA, HVGA                                                                                                                                                                                                                                                                                                                                                                  | Under study                   |
+| Coded video bitrate            | QCIF: 50-1000 kbit/s<br>QVGA: 80-3000 kbit/s<br>HVGA: 192-6000 kbit/s                                                                                                                                                                                                                                                                                                             | Under study                   |
+
+**Table 4 – Factors and application ranges of the ITU-T P.1202 model algorithms**
+
+|                                                                                                      | <b>ITU-T P.1202.1 lower resolution (LR)</b>                                                                                                                                                                                                                                                                                                                                                                                      | <b>Higher resolution (HR)</b> |
+|------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------|
+| <b>Application information</b>                                                                       | <b>Value range, unit</b>                                                                                                                                                                                                                                                                                                                                                                                                         |                               |
+| Video decoder packet loss concealment                                                                | Two types of assumed decoder behaviour are covered:<br>1) freezing with skipping,<br>2) slicing with:<br>ITU-T H.264: 1 slice/packet.<br>Fixed PLC (using fixed decoder, details and settings)                                                                                                                                                                                                                                   | Under study                   |
+| Retransmission mechanisms (ARQ);<br>Forward Error Correction (FEC)<br>Client jitter buffer behaviour | Rebuffering handling, particular to LR-case: without skipping of length 0 to 8 seconds<br>Developed models represent CC-mode (see clause 6.1), hence applied as if dejitter buffer, ARQ and FEC mechanisms have already corrected the stream<br>For other modes of operation, see clause 6.1, appropriate methods to correct the streams in ways reflecting the expected FEC, ARQ and dejitter buffer behaviour are under study. | Under study                   |
+| Encoder implementation                                                                               | The model has been trained using the following video encoders (Note 1):<br>– ITU-T H.264/AVC: x264                                                                                                                                                                                                                                                                                                                               | Under study                   |
+| Decoder implementation                                                                               | Reference decoder was a proprietary decoder provided by one proponent, which also performed de-packetization and audio/video-demultiplexing. The ITU-T H.264-decoding is standard-conformant, with the PLC as described above (Note 2).                                                                                                                                                                                          | Under study                   |
+| Group of pictures (GOP)                                                                              | GOP-structure is estimated from the stream.<br>Typical GOP structure for which the model has been trained:<br>M = 1, N = 40 (typically no B frames for mobile case)<br>Length: fixed, variable, adaptive<br>Structure (e.g., IPPP..PPPI)                                                                                                                                                                                         | Under study                   |
+| Frame rate                                                                                           | 12.5, 15, 20, 25, 30 fps                                                                                                                                                                                                                                                                                                                                                                                                         | Under study                   |
+| Usage of: Marker bit in RTP header                                                                   | "End of frame" (True/false)                                                                                                                                                                                                                                                                                                                                                                                                      | Under study                   |
+| Encrypted payload                                                                                    | Not applicable                                                                                                                                                                                                                                                                                                                                                                                                                   | Not applicable                |
+| Packet loss degradation, video                                                                       | Uniform loss:<br>0-6%<br>Burst loss:<br>0-6% (4-state Markov model)                                                                                                                                                                                                                                                                                                                                                              | Under study                   |
+
+**Table 4 – Factors and application ranges of the ITU-T P.1202 model algorithms**
+
+|                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             | <b>ITU-T P.1202.1 lower resolution (LR)</b> | <b>Higher resolution (HR)</b> |
+|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------|-------------------------------|
+| <b>Application information</b>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              | <b>Value range, unit</b>                    |                               |
+| <p>NOTE 1 – It is assumed that the model can be used for estimating quality when other encoder implementations for the given codec have been used. However, if the encoder performance is significantly worse or better than for the encoder used, the model prediction accuracy will be reduced.</p> <p>NOTE 2 – One aspect not covered by decoder packet loss concealment is postfiltering. Guidance on how to adjust internal model parameters for specific other decoders incl. set-top boxes is for further study.</p> |                                             |                               |
+
+## 6.2 Modes of operation
+
+The four modes of operation are described in Table 5 and Figures 1-a to 1-e below. Note that the model as described in [ITU-T P.1202.1] support one of the four possible modes (the so-called CC mode). Additional adaptation is required to use the ITU-T P.1202 for the other modes.
+
+**Table 5 – Modes of operations of ITU-T P.1202**
+
+| <b>Class</b>                                                                                                                                                                                                     | <b>Name</b>                        | <b>Mode abbreviation*</b> | <b>Description</b>                                                                                                                                                                                                                                            |
+|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------|---------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| Mid-point or End-point                                                                                                                                                                                           | Static operation                   | NN                        | The model uses information from the local transport layer, prior knowledge about coding and prior knowledge about the end-point                                                                                                                               |
+| Mid-point                                                                                                                                                                                                        | Non-embedded dynamic operation     | BN                        | The model uses information from the local transport layer, prior knowledge about coding and information about the end-point collected through measurement reporting protocols                                                                                 |
+| Mid-point                                                                                                                                                                                                        | Non-embedded distributed operation | CN                        | The model, located inside the network, uses information from the transport layer measured at an end-point and collected through signalling protocols, prior knowledge about coding and information about the end-point collected through signalling protocols |
+| End-point                                                                                                                                                                                                        | Embedded operation                 | CC                        | The model uses information from the local transport layer, information from the end-point, and prior knowledge about coding                                                                                                                                   |
+| <p>*Mode abbreviation naming scheme:<br/>XY, where<br/>X corresponds to place of measurement (N: Network, C: Client, B: Both network and client)<br/>Y corresponds to place of model (N: Network, C: Client)</p> |                                    |                           |                                                                                                                                                                                                                                                               |
+
+In Figures 1-a to 1-e below the following arrow style is used:
+
+- Media stream
+- - - -> Signalling protocol
+- - - -> Static information
+- . - . -> Buffering information
+
+![Figure 1-a: Static operation mode (NN) inside the network. A 'Send point' block sends a 'Media stream' (solid line) to an 'End point' block, which outputs a 'Media signal'. A 'Buffering estimation' block receives a 'Buffering information' (dash-dot line) from the media stream and sends 'Static information' (dashed line) to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.](724c7777b608e53be38b12b6fb3c43bc_img.jpg)
+
+Figure 1-a: Static operation mode (NN) inside the network. A 'Send point' block sends a 'Media stream' (solid line) to an 'End point' block, which outputs a 'Media signal'. A 'Buffering estimation' block receives a 'Buffering information' (dash-dot line) from the media stream and sends 'Static information' (dashed line) to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.
+
+Figure 1-a – Static operation mode (NN) inside the network
+
+![Figure 1-b: Static operation mode (NN) inside a terminal. A 'Send point' block sends a 'Media stream' (solid line) to a 'Media buffer' block inside an 'End point' (dashed box). The 'Media buffer' feeds into a 'Decoder and PLC' block, which outputs a 'Media signal'. A 'Buffering estimation' block receives 'Buffering information' (dash-dot line) from the media stream and sends 'Static information' (dashed line) to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.](75f0cb39f1cd165dfe4a6aa6c4d9388d_img.jpg)
+
+Figure 1-b: Static operation mode (NN) inside a terminal. A 'Send point' block sends a 'Media stream' (solid line) to a 'Media buffer' block inside an 'End point' (dashed box). The 'Media buffer' feeds into a 'Decoder and PLC' block, which outputs a 'Media signal'. A 'Buffering estimation' block receives 'Buffering information' (dash-dot line) from the media stream and sends 'Static information' (dashed line) to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.
+
+Figure 1-b – Static operation mode (NN) inside a terminal
+
+![Figure 1-c: Non-embedded dynamic operation mode (BN). A 'Send point' block sends a 'Media stream' (solid line) to an 'End point' block, which outputs a 'Media signal'. A 'Buffering information' (dash-dot line) is sent from the 'End point' to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.](9ae17964ddd9b814c7d905b1af2fddf2_img.jpg)
+
+Figure 1-c: Non-embedded dynamic operation mode (BN). A 'Send point' block sends a 'Media stream' (solid line) to an 'End point' block, which outputs a 'Media signal'. A 'Buffering information' (dash-dot line) is sent from the 'End point' to a 'Model' block. The 'Model' block also receives 'Static media stream and decoder behaviour information' (dashed line) from a cylinder icon and outputs a 'MOS' value.
+
+Figure 1-c – Non-embedded dynamic operation mode (BN)
+
+![Figure 1-d: Non-embedded distributed operation mode (CN) diagram](fa859e4e468bfb2710a94527f2c504af_img.jpg)
+
+This diagram illustrates the non-embedded distributed operation mode (CN). A 'Send point' block on the left sends a 'Media stream' (solid blue arrow) to an 'End point' block on the right. The 'End point' outputs a 'Media signal' (solid blue arrow). A dashed blue arrow from the 'End point' points down to a 'Model' block. A cylinder labeled 'Static media stream and decoder behaviour information' also points to the 'Model' block via a dashed blue arrow. The 'Model' block outputs a 'MOS' (Mean Opinion Score) value (solid blue arrow).
+
+Figure 1-d: Non-embedded distributed operation mode (CN) diagram
+
+**Figure 1-d – Non-embedded distributed operation mode (CN)**
+
+![Figure 1-e: Embedded operation mode (CC) diagram](0f985b39edc1d52ba3600c438bc8f0a5_img.jpg)
+
+This diagram illustrates the embedded operation mode (CC). A 'Send point' block on the left sends a 'Media stream' (solid blue arrow) into a 'Media buffer' block. The 'Media buffer' block is part of an 'End point' (indicated by a dashed blue box). From the 'Media buffer', a solid blue arrow points to a 'Decoder and PLC' block, which outputs a 'Media signal' (solid blue arrow). A dashed blue arrow from the 'Media buffer' points up to a 'Model' block. A cylinder labeled 'Static media stream and decoder behaviour information' also points to the 'Model' block via a dashed blue arrow. The 'Model' block outputs a 'MOS' value (solid blue arrow).
+
+Figure 1-e: Embedded operation mode (CC) diagram
+
+**Figure 1-e – Embedded operation mode (CC)**
+
+# 7 Model input interfaces
+
+The ITU-T P.1202 model will receive the encoded bitstream and static side information. For the model as it is described in [ITU-T P.1202.1], the encoded bitstream is expected to be provided in a PCAP file format with transport header information. However, in practical implementations, other than PCAP-based realizations of the transport layer packet extraction can be envisaged. For models as they are described here, the PCAP file could be created based on packets being captured at a network interface. The static side information is information about the media stream and the decoder behaviour. The overview information per application area and mode is described in Table 6 and Figure 2.
+
+![Figure 2 – Overview of ITU-T P.1202 model interfaces. The diagram shows the flow of data through the P.1202 model. An 'Encoded bit stream' enters from the left via interface I.2. Inside the 'P.1202 model' (dashed box), the stream goes to 'Parameter extraction', which outputs via interface I.3 to 'MOS estimation (model core)'. Below the model, 'Static side information' (cylinder) connects via interface I.1 and 'Buffering information' (box) connects via interface I.4 to the 'MOS estimation' block. The final output is 'MOS'.](8fbdfc3d17fb1dae7b2d8f5a287fa9fc_img.jpg)
+
+```
+
+graph LR
+    EBS[Encoded bit stream] -- I.2 --> PE[Parameter extraction]
+    subgraph P1202_model [P.1202 model]
+        PE -- I.3 --> MEC[MOS estimation (model core)]
+    end
+    SSI[(Static side information)] -- I.1 --> MEC
+    BI[Buffering information] -- I.4 --> MEC
+    MEC -- MOS --> MOS_out[MOS]
+  
+```
+
+Figure 2 – Overview of ITU-T P.1202 model interfaces. The diagram shows the flow of data through the P.1202 model. An 'Encoded bit stream' enters from the left via interface I.2. Inside the 'P.1202 model' (dashed box), the stream goes to 'Parameter extraction', which outputs via interface I.3 to 'MOS estimation (model core)'. Below the model, 'Static side information' (cylinder) connects via interface I.1 and 'Buffering information' (box) connects via interface I.4 to the 'MOS estimation' block. The final output is 'MOS'.
+
+**Figure 2 – Overview of ITU-T P.1202 model interfaces**
+
+The I.3 interface in the CN mode conveys the signalling information from the parameter extraction module located in the end-point.
+
+The ITU-T P.1202 model has three main inputs:
+
+- Encoded video bitstream: This input can be extracted from a PCAP file, or streamed from a network interface. Parameters from the bitstream can be extracted from the bitstream in the end-point of a CN mode implementation and sent back to the MOS estimation (model core) located in the "model" node.
+- Buffering information: This input taken from the media buffer in the client, or estimated by a buffering estimation module using packet information.
+- Static media- and decoder information: This input is obtained from packet information or from a player API.
+
+**Table 6 – Overview of input to the ITU-T P.1202 model for the modes of operation**
+
+|                                                                                 | <b>Static operation (NN)</b>                                                                                                                      | <b>Non-embedded dynamic operation (BN)</b>                                                                                                                                                                                                            | <b>Non-embedded distributed operation (CN)</b>                                                                                                                                                                                                     | <b>Embedded operation (CC)</b>                                                                                                                 |
+|---------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------|
+| <b>Interface 1 (I.1)</b>                                                        | Static information about the media stream (codec, usage of flags etc.), and static information about the decoder. Detailed description in Table 7 | Static information about the media stream (codec, usage of flags etc.), and static information about the decoder. Detailed description in Table 7<br>NOTE – Static decoder behaviour information might be provided via signalling and included in I.3 | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7<br>NOTE – Static decoder behaviour information might be provided via signalling and included in I.3 | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7 |
+| <b>Interface 2 (I.2)</b>                                                        | PCAP file (payload required)                                                                                                                      | PCAP file (payload required)                                                                                                                                                                                                                          | PCAP file (payload required)                                                                                                                                                                                                                       | PCAP file (payload required)                                                                                                                   |
+| <b>Interface 3 (I.3)</b>                                                        | Parameters extracted from the PCAP file                                                                                                           | Output from a parameter extraction module located in the end-point and transferred to the model with a protocol                                                                                                                                       | Output from a parameter extraction module located in the end-point and transferred to the model with a protocol                                                                                                                                    | Parameters extracted from the PCAP file                                                                                                        |
+| <b>Interface 4 (I.4)</b><br><b>Only available for the ITU-T P.1202 LR model</b> | Rebuffering information (estimated)                                                                                                               | Rebuffering parameters measured/extracted in the end-point and transferred to the model with a protocol                                                                                                                                               | Rebuffering parameters measured/extracted in the end-point and transferred to the model with a protocol                                                                                                                                            | Rebuffering information from the media buffer                                                                                                  |
+
+Table 7 gives more detailed examples of information that is provided at the input interfaces I.1 (see Figure 2).
+
+**Table 7 – Input and typical values to ITU-T P.1201 models**
+
+| <b>Input</b>              | <b>Typical value</b>                         |
+|---------------------------|----------------------------------------------|
+| <b>Dynamic input</b>      |                                              |
+| Media stream              | PCAP file or other capture format            |
+| Rebuffering information   | Text file containing rebuffering information |
+| <b>Static information</b> |                                              |
+| Video destination port    | 1234                                         |
+| Video codec               | H264                                         |
+| Video codec profile       | Baseline                                     |
+| Video resolution          | QCIF, QVGA, HVGA                             |
+
+**Table 7 – Input and typical values to ITU-T P.1201 models**
+
+| Input                         | Typical value     |
+|-------------------------------|-------------------|
+| Video scanning type           | Progressive       |
+| Video frame rate              | 12.5, 15, 25, 30  |
+| Video packet loss concealment | Slicing, freezing |
+
+The models described in ITU-T P.1202 have been validated assuming that the available information already reflects the impact of any error resilience methods, such as forward error correction (FEC), or packet re-transmission mechanisms such as automatic repeat request (ARQ), and of the impact due to the dejitter buffer. This is equivalent to the parameter extraction module being located behind these processing steps, that is, implementing the CC mode of operation. In case of the NN mode, the measurement point is located prior to the actual FEC, ARQ and dejitter buffer mechanisms. Since these mechanisms may have a very strong impact on factors such as the packet loss seen by the decoder, the NN-mode requires an explicit handling of these mechanisms, to reflect a CN or BN type of behaviour. To capture this case, the packet stream may be converted into a stream that reflects an assumed behaviour of FEC, ARQ, and/or jitter de-buffering, reflecting the input format to be provided to the parameter extraction module. This step results in a converted stream, see Figure 3-b.
+
+Figure 3 shows how the case of error resilience methods such as forward error correction (FEC) and automatic repeat request (ARQ) could be handled in case of the NN/CN/BN modes. Figure 3-a shows the model architecture when FEC/ARQ is not used. In Figure 3-b the media stream is first corrected using a FEC/ARQ dejitter buffer and then the parameters are extracted in exactly the same way as in the case without FEC/ARQ.
+
+![Block diagram of the model structure when error resilience methods (FEC/ARQ) are not used. A 'Media stream without FEC/ARQ' enters a 'Parameter extraction' block. This block outputs 'Internal parameters' to a 'Model core' block. The 'Model core' block also receives 'Static media stream and decoder behaviour information' from a database (cylinder icon) and outputs the final 'MOS' score.](10781f43062bf3e9601a1e086710556c_img.jpg)
+
+```
+graph LR; A[Media stream without FEC/ARQ] --> B[Parameter extraction]; B -- Internal parameters --> C[Model core]; D[(Static media stream and decoder behaviour information)] -.-> C; C --> E[MOS]
+```
+
+Block diagram of the model structure when error resilience methods (FEC/ARQ) are not used. A 'Media stream without FEC/ARQ' enters a 'Parameter extraction' block. This block outputs 'Internal parameters' to a 'Model core' block. The 'Model core' block also receives 'Static media stream and decoder behaviour information' from a database (cylinder icon) and outputs the final 'MOS' score.
+
+**Figure 3-a – Model structure when error resilience methods (FEC/ARQ) are not used**
+
+![Figure 3-b: Model structure when FEC/ARQ is used and the stream is corrected before parameter extraction. The flowchart shows: Media stream using FEC/ARQ -> FEC/ARQ de-jitter buffer -> Corrected media stream -> Parameter extraction -> Internal parameters -> Model core -> MOS. A database labeled 'Static media stream and decoder behaviour information' is connected to the Model core via a dashed arrow.](b6671cfafda3820aafe9a24fa7a4d8c7_img.jpg)
+
+```
+
+graph LR
+    A[Media stream using FEC/ARQ] --> B[FEC/ARQ de-jitter buffer]
+    B --> C[Corrected media stream]
+    C --> D[Parameter extraction]
+    D --> E[Internal parameters]
+    E --> F[Model core]
+    F --> G[MOS]
+    H[(Static media stream and decoder behaviour information)] -.-> F
+  
+```
+
+Figure 3-b: Model structure when FEC/ARQ is used and the stream is corrected before parameter extraction. The flowchart shows: Media stream using FEC/ARQ -> FEC/ARQ de-jitter buffer -> Corrected media stream -> Parameter extraction -> Internal parameters -> Model core -> MOS. A database labeled 'Static media stream and decoder behaviour information' is connected to the Model core via a dashed arrow.
+
+**Figure 3-b – Model structure when FEC/ARQ is used and the stream is corrected before parameter extraction**
+
+# 8 Model output information and performance details
+
+The ITU-T P.1202 models have one output parameter:
+
+- Estimated video MOS on the 1 to 5 scale, which is an estimation of the perceived video quality.
+
+The performance information for the ITU-T P.1202.1 model can be found in Table 8 and in Appendix I. The performance information for the ITU-T P.1202.2 models can be found in Table 9, Table 10, Appendix II and Appendix III. The statistical metrics RMSE (root mean square error) and Pearson correlation are used to describe the performance, see [ITU-T P.1401]. Note that for those performance figures, the subjective ratings have been mapped to the model scores using a linear, i.e., 1st-order mapping function, at a per-database level. This has been done in order to avoid misalignment due to bias in the different subjective tests, e.g., as a result of different test settings.
+
+**Table 8 – Performance information for ITU-T P.1202.1**
+
+|                     | RMSE                           | Pearson correlation            |
+|---------------------|--------------------------------|--------------------------------|
+| Overall performance | 0.397 (based on 982 sequences) | 0.918 (based on 982 sequences) |
+
+**Table 9 – Performance information for ITU-T P.1202.2 mode 1**
+
+|                     | RMSE                            | Pearson correlation             |
+|---------------------|---------------------------------|---------------------------------|
+| Overall performance | 0.357 (based on 3069 sequences) | 0.938 (based on 3069 sequences) |
+
+**Table 10 – Performance information for ITU-T P.1202.2 mode 2**
+
+|                     | RMSE                            | Pearson correlation             |
+|---------------------|---------------------------------|---------------------------------|
+| Overall performance | 0.353 (based on 3069 sequences) | 0.940 (based on 3069 sequences) |
+
+# **9 Description of the ITU-T P.1202 model algorithm**
+
+The ITU-T P.1202 lower resolution model is described in [ITU-T P.1202.1]. The ITU-T P.1202 higher resolution model is described in [ITU-T P.1202.2].
+
+## Appendix I
+
+### Detailed performance figures for the ITU-T P.1202.1 algorithm
+
+(This appendix does not form an integral part of this Recommendation.)
+
+|                                                        | <b>ITU-T P.1202.1 (lower resolution)</b> |
+|--------------------------------------------------------|------------------------------------------|
+| Overall video RMSE                                     | 0.397 (based on 982 sequences)           |
+| RMSE for video packet loss conditions causing freezing | 0.400 (based on 104 sequences)           |
+| RMSE for video packet loss conditions causing slicing  | 0.508 (based on 422 sequences)           |
+| RMSE for video rebuffering                             | 0.299 (based on 168 sequences)           |
+| RMSE for pure compression conditions                   | 0.284 (based on 288 sequences)           |
+| Pearson correlation                                    | 0.918 (based on 982 sequences)           |
+
+| <b>Media</b> | <b>Codec</b> | <b>Degradation type</b>                     | <b>RMSE</b> | <b>PC</b> | <b># files</b> |
+|--------------|--------------|---------------------------------------------|-------------|-----------|----------------|
+| Video        | Overall      |                                             | 0.397       | 0.918     | 982            |
+|              | H264 (QCIF)  | Compression, Slicing, Freezing, Rebuffering | 0.442       | 0.895     | 207            |
+|              | H264 (QVGA)  | Compression, Slicing, Freezing, Rebuffering | 0.365       | 0.931     | 375            |
+|              | H264 (HVGA)  | Compression, Slicing, Freezing, Rebuffering | 0.402       | 0.921     | 400            |
+
+## Appendix II
+
+### Detailed performance figures for the ITU-T P.1202.2 mode 1 algorithm
+
+(This appendix does not form an integral part of this Recommendation.)
+
+|                                                        | <b>ITU-T P.1202.2 (higher resolution) mode 1</b> |
+|--------------------------------------------------------|--------------------------------------------------|
+| Overall video RMSE                                     | 0.357 (based on 3069 sequences)                  |
+| RMSE for video packet loss conditions causing freezing | 0.313 (based on 687 sequences)                   |
+| RMSE for video packet loss conditions causing slicing  | 0.396 (based on 1374 sequences)                  |
+| RMSE for pure compression conditions                   | 0.325 (based on 1008 sequences)                  |
+| Pearson correlation                                    | 0.938 (based on 3069 sequences)                  |
+
+| <b>Media</b> | <b>Codec</b> | <b>Degradation type</b>        | <b>RMSE</b> | <b>PC</b> | <b># files</b> |
+|--------------|--------------|--------------------------------|-------------|-----------|----------------|
+| Video        |              | Overall                        | 0.357       | 0.938     | 3069           |
+|              | H264 (SD)    | Compression, Slicing, Freezing | 0.337       | 0.940     | 698            |
+|              | H264 (720P)  | Compression, Slicing, Freezing | 0.354       | 0.938     | 719            |
+|              | H264 (HD)    | Compression, Slicing, Freezing | 0.368       | 0.937     | 1652           |
+
+## Appendix III
+
+### Detailed performance figures for the ITU-T P.1202.2 mode 2 algorithm
+
+(This appendix does not form an integral part of this Recommendation.)
+
+|                                                        | <b>ITU-T P.1202.2 (higher resolution) mode 2</b> |
+|--------------------------------------------------------|--------------------------------------------------|
+| Overall video RMSE                                     | 0.353 (based on 3069 sequences)                  |
+| RMSE for video packet loss conditions causing freezing | 0.287 (based on 687 sequences)                   |
+| RMSE for video packet loss conditions causing slicing  | 0.392 (based on 1374 sequences)                  |
+| RMSE for pure compression conditions                   | 0.337 (based on 1008 sequences)                  |
+| Pearson correlation                                    | 0.940 (based on 3069 sequences)                  |
+
+| <b>Media</b> | <b>Codec</b> | <b>Degradation type</b>        | <b>RMSE</b> | <b>PC</b> | <b># files</b> |
+|--------------|--------------|--------------------------------|-------------|-----------|----------------|
+| Video        | Overall      |                                | 0.353       | 0.940     | 3069           |
+|              | H264 (SD)    | Compression, Slicing, Freezing | 0.335       | 0.943     | 698            |
+|              | H264 (720P)  | Compression, Slicing, Freezing | 0.346       | 0.942     | 719            |
+|              | H264 (HD)    | Compression, Slicing, Freezing | 0.364       | 0.937     | 1652           |
+
+
+
+## SERIES OF ITU-T RECOMMENDATIONS
+
+|                 |                                                                                             |
+|-----------------|---------------------------------------------------------------------------------------------|
+| Series A        | Organization of the work of ITU-T                                                           |
+| Series D        | General tariff principles                                                                   |
+| Series E        | Overall network operation, telephone service, service operation and human factors           |
+| Series F        | Non-telephone telecommunication services                                                    |
+| Series G        | Transmission systems and media, digital systems and networks                                |
+| Series H        | Audiovisual and multimedia systems                                                          |
+| Series I        | Integrated services digital network                                                         |
+| Series J        | Cable networks and transmission of television, sound programme and other multimedia signals |
+| Series K        | Protection against interference                                                             |
+| Series L        | Construction, installation and protection of cables and other elements of outside plant     |
+| Series M        | Telecommunication management, including TMN and network maintenance                         |
+| Series N        | Maintenance: international sound programme and television transmission circuits             |
+| Series O        | Specifications of measuring equipment                                                       |
+| <b>Series P</b> | <b>Terminals and subjective and objective assessment methods</b>                            |
+| Series Q        | Switching and signalling                                                                    |
+| Series R        | Telegraph transmission                                                                      |
+| Series S        | Telegraph services terminal equipment                                                       |
+| Series T        | Terminals for telematic services                                                            |
+| Series U        | Telegraph switching                                                                         |
+| Series V        | Data communication over the telephone network                                               |
+| Series X        | Data networks, open system communications and security                                      |
+| Series Y        | Global information infrastructure, Internet protocol aspects and next-generation networks   |
+| Series Z        | Languages and general software aspects for telecommunication systems                        |

marked/P/T-REC-P.1203.2-201710-I_PDF-E/raw.md

@@ -0,0 +1,387 @@
+
+
+International Telecommunication Union
+
+**ITU-T**
+
+**P.1203.2**
+
+TELECOMMUNICATION  
+STANDARDIZATION SECTOR  
+OF ITU
+
+(10/2017)
+
+SERIES P: TELEPHONE TRANSMISSION QUALITY,  
+TELEPHONE INSTALLATIONS, LOCAL LINE  
+NETWORKS
+
+Models and tools for quality assessment of streamed  
+media
+
+---
+
+**Parametric bitstream-based quality assessment  
+of progressive download and adaptive  
+audiovisual streaming services over reliable  
+transport – Audio quality estimation module**
+
+Recommendation ITU-T P.1203.2
+
+ITU-T
+
+![ITU logo](1d7527f4316cfe2d342b08d1653d1592_img.jpg)
+
+The logo of the International Telecommunication Union (ITU) features a stylized globe with a red lightning bolt striking across it. The letters 'ITU' are prominently displayed in blue and red.
+
+ITU logo
+
+International  
+Telecommunication  
+Union
+
+# ITU-T P-SERIES RECOMMENDATIONS
+
+## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS
+
+|                                                                                                    |               |               |
+|----------------------------------------------------------------------------------------------------|---------------|---------------|
+| Vocabulary and effects of transmission parameters on customer opinion of transmission quality      | Series        | P.10          |
+| Voice terminal characteristics                                                                     | Series        | P.30          |
+|                                                                                                    |               | P.300         |
+| Reference systems                                                                                  | Series        | P.40          |
+| Objective measuring apparatus                                                                      | Series        | P.50          |
+|                                                                                                    |               | P.500         |
+| Objective electro-acoustical measurements                                                          | Series        | P.60          |
+| Measurements related to speech loudness                                                            | Series        | P.70          |
+| Methods for objective and subjective assessment of speech quality                                  | Series        | P.80          |
+| Methods for objective and subjective assessment of speech and video quality                        | Series        | P.800         |
+| Audiovisual quality in multimedia services                                                         | Series        | P.900         |
+| Transmission performance and QoS aspects of IP end-points                                          | Series        | P.1000        |
+| Communications involving vehicles                                                                  | Series        | P.1100        |
+| <b>Models and tools for quality assessment of streamed media</b>                                   | <b>Series</b> | <b>P.1200</b> |
+| Telemeeting assessment                                                                             | Series        | P.1300        |
+| Statistical analysis, evaluation and reporting guidelines of quality measurements                  | Series        | P.1400        |
+| Methods for objective and subjective assessment of quality of services other than speech and video | Series        | P.1500        |
+
+For further details, please refer to the list of ITU-T Recommendations.
+
+## Recommendation ITU-T P.1203.2
+
+## Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module
+
+## Summary
+
+Recommendation ITU-T P.1203.2 specifies the short-term audio quality estimation module for Recommendation ITU-T P.1203. The ITU-T P.1203 series of ITU-T Recommendations specifies modules for a set of model algorithms for monitoring the integral media session quality for transport control protocol (TCP) type video streaming. The models comprise modules for short-term video-quality and audio-quality estimation (the latter specified in this Recommendation). The per-one-second outputs of these short-term modules are integrated into estimates of audio-visual quality and together with information about initial loading delay and media playout stalling events, they are further integrated into the final model output, the estimate of integral quality. The respective ITU-T work item has formerly been referred to as "Parametric non-intrusive assessment of TCP-based multimedia streaming quality" or "P.NATS". The Recommendation ITU-T P.1203.2 part of Recommendation ITU-T P.1203 provides details for the module for bitstream-based, short-term audio quality estimation.
+
+Only one audio module is recommended for all four modes 0 to 3 of the Recommendation ITU-T P.1203 model series, corresponding to mode 0. The model is identical to the audio coding quality estimation component of the user datagram protocol (UDP) streaming related prediction model described in Recommendation ITU-T P.1201.
+
+## History
+
+| Edition | Recommendation | Approval   | Study Group | Unique ID*                                                                |
+|---------|----------------|------------|-------------|---------------------------------------------------------------------------|
+| 1.0     | ITU-T P.1203.2 | 2016-11-29 | 12          | <a href="http://handle.itu.int/11.1002/1000/13160">11.1002/1000/13160</a> |
+| 2.0     | ITU-T P.1203.2 | 2017-10-29 | 12          | <a href="http://handle.itu.int/11.1002/1000/13401">11.1002/1000/13401</a> |
+
+## Keywords
+
+Adaptive streaming, audio, audiovisual, IPTV, mean opinion score (MOS), mobile video, mobile TV, monitoring, multimedia, progressive download, QoE, TV, video.
+
+---
+
+\* To access the Recommendation, type the URL <http://handle.itu.int/> in the address field of your web browser, followed by the Recommendation's unique ID. For example, <http://handle.itu.int/11.1002/1000/11830-en>.
+
+## FOREWORD
+
+The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis.
+
+The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics.
+
+The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1.
+
+In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC.
+
+## NOTE
+
+In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency.
+
+Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party.
+
+## INTELLECTUAL PROPERTY RIGHTS
+
+ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process.
+
+As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at <http://www.itu.int/ITU-T/ipr/>.
+
+© ITU 2017
+
+All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU.
+
+## Table of Contents
+
+|                                               | Page |
+|-----------------------------------------------|------|
+| 1 Scope.....                                  | 1    |
+| 2 References.....                             | 3    |
+| 3 Definitions .....                           | 3    |
+| 3.1 Terms defined elsewhere .....             | 3    |
+| 3.2 Terms defined in this Recommendation..... | 4    |
+| 4 Abbreviations and acronyms .....            | 4    |
+| 5 Conventions .....                           | 4    |
+| 6 Pa module in ITU-T P.1203 context.....      | 4    |
+| 6.1 <i>Pa</i> module modes .....              | 5    |
+| 7 Model input.....                            | 5    |
+| 7.1 I.11 input specification .....            | 6    |
+| 8 Model algorithm and output .....            | 6    |
+| Bibliography.....                             | 8    |
+
+
+
+## Recommendation ITU-T P.1203.2
+
+### Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module
+
+# 1 Scope
+
+This Recommendation describes the short term audio quality estimation module which is an integral part of the ITU-T P.1203 series. [ITU-T P.1203] describes a set of objective parametric quality assessment modules. Combined, these modules can be used to predict the impact of audio and video media encodings as well as Internet protocol (IP) network impairments on the quality experienced by an end-user of multi-media streaming applications.
+
+The addressed streaming techniques comprise progressive download as well as adaptive streaming, for both mobile and fixed network streaming applications over transport control protocol (TCP) or other TCP like protocols which are not affected by transmission errors.
+
+The model described is restricted to information provided to it by an appropriate packet- or bitstream-analysis module. The overall ITU-T P.1203 model is applicable for the effects due to audio- and video-coding as well as initial loading delay and stalling (which are both caused by rebuffering at the client) as the typical degradations associated with progressive download. As final output, the ITU-T P.1203 series models target integral audio-visual media quality scores.
+
+This Recommendation describes only one audio quality module. With regard to the required input data, this audio module corresponds to mode 0 of [ITU-T P.1203].
+
+The same, purely header-based/bitrate-based audio quality module is also specified in [ITU-T P.1201.2]. Using a large number of subjective experiments, it was validated that this model also leads to accurate predictions within the scope of [ITU-T P.1203].
+
+The audio module predicts mean opinion scores (MOS) on a 5-point absolute category rating (ACR) scale (see [ITU-T P.910]) as a per-one-second MOS score.
+
+During the development of [ITU-T P.1201], explicit short-term audio quality tests were carried out in order to validate the stand-alone use of the audio module for the estimation of audio-only quality. It could be shown within the scope of [ITU-T P.1201] that this is possible.
+
+It must be noted however, that since the subjective tests conducted for [ITU-T P.1201] included packet loss degradations, range-equalization and other biases may need to be considered (see for example [b-Zielinski\_2008]) if the module is to be used stand-alone within the scope of [ITU-T P.1203].
+
+This model cannot provide a comprehensive evaluation of audio transmission quality as perceived by an *individual* end user because its scores reflect the impairments due to audio coding only. Furthermore, the scores predicted by a parametric model necessarily reflect an average perceptual impairment. Note also that the model was developed and validated for one specific encoder and decoder implementation. If a different encoder and decoder pair is used in a monitoring situation the scores may not reflect that.
+
+Effects such as audio level or noise (and corresponding similar audio factors) or other impairments related to the audio signals are not reflected in the scores computed by this model. Moreover, the scores predicted by a parametric model (i.e., without access to payload information, such as the audio signals) necessarily reflect a somewhat simplified representation of the perceptual impairment of the considered stream.
+
+However, presuming that it is applied in an appropriate manner, according to this Recommendation, the model still enables estimation of some coding quality related information and thus valid and in most cases accurate predictions.
+
+Tables 1.1 and 1.2 indicate the areas and parameter ranges for which the Pa module specified in this Recommendation has been validated and for which applications it can be used, with some caution.
+
+**Table 1.1 – Application areas, test factors and coding technologies where ITU-T P.1203.2 for adaptive streaming and progressive download has been verified and is known to produce reliable results**
+
+| <b>Applications for which the model is intended</b>                                                                                                                                                                                                                                                                  |                                                                                                                                                                                                                                                                                                                                  |
+|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| In-service monitoring of TCP-based audio. Both so called over the top (OTT) services (for example YouTube) and operator managed video services (over TCP), using the protocols HTTP/TCP/IP and RTMP/TCP/IP.<br>Note that this model is agnostic to the type of container format (e.g. Flash (FLV), MP4, WebM or 3GP. |                                                                                                                                                                                                                                                                                                                                  |
+| Performance and quality assessment of live networks (including codecs) considering the effect due to encoding bit rate.                                                                                                                                                                                              |                                                                                                                                                                                                                                                                                                                                  |
+| <b>Audio test factors for which the model has been validated</b>                                                                                                                                                                                                                                                     |                                                                                                                                                                                                                                                                                                                                  |
+| Input audio length                                                                                                                                                                                                                                                                                                   | Maximum 20 seconds. The video model produces a per-second score considering input data from a measurement window of max. 20 s length.                                                                                                                                                                                            |
+| Bitstream container                                                                                                                                                                                                                                                                                                  | Coded audio bitstream contained in MPEG-2 transport stream (TS) segments                                                                                                                                                                                                                                                         |
+| Encoder/Decoder implementation                                                                                                                                                                                                                                                                                       | The model has been trained using the following audio encoder: <ul style="list-style-type: none"> <li>– AAC-LC: libfdk_aac, low complexity (LC) mode (ffmpeg).</li> <li>– A common framework was developed based on the above codec, all the test data was generated using the common framework.</li> </ul>                       |
+| Audio sample rate                                                                                                                                                                                                                                                                                                    | 48 000 samples/s                                                                                                                                                                                                                                                                                                                 |
+| Audio bit rate                                                                                                                                                                                                                                                                                                       | 16, 32, 64 and 98 kBit/s/channel<br>Audio bit rate was always varied in a correlated fashion with the video bit rate, i.e., high video bit rate corresponds to high audio bit rate and vice versa. Bearing to this condition it has been observed that audio quality has very little effect on the overall audio-visual quality. |
+| Segment length                                                                                                                                                                                                                                                                                                       | 1-9 seconds<br>NOTE – The segment length determines how often the audio quality can be adapted.                                                                                                                                                                                                                                  |
+| Audio channels                                                                                                                                                                                                                                                                                                       | 2 (stereo)                                                                                                                                                                                                                                                                                                                       |
+
+**Table 1.2 – Application areas, test factors and coding technologies for which ITU-T P.1203.2 is assumed to give valid results**
+
+| <b>Test factors where the model can be used but the results may not be reliable (conditions not included in subjective tests underlying the model development)</b>                          |
+|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| All factors as indicated in Table 1.1, with additions as described below:<br>Codecs: HE-AACv2, AC3, MPEG-LII<br>Bit rates: 4.75-576 kbit/s                                                  |
+| NOTE – ITU-T P.1203 was tested on AAC-LC only. The audio module alone has been tested with the codecs mentioned above with dedicated audio-quality tests during [ITU-T P.1201] development. |
+
+# 2 References
+
+The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation.
+
+- [ITU-T P.800.1] Recommendation ITU-T P.800.1 (2016), *Mean opinion score (MOS) terminology*.
+- [ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*.
+- [ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*.
+- [ITU-T P.1201] Recommendation ITU-T P.1201 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality*.
+- [ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Lower resolution application area*.
+- [ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Higher resolution application area*.
+- [ITU-T P.1202] Recommendation ITU-T P.1202 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality*.
+- [ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – Lower resolution application area*.
+- [ITU-T P.1203] Recommendation ITU-T P.1203 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport*.
+- [ITU-T P.1203.1] Recommendation ITU-T P.1203.1 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport –Video quality estimation module*.
+- [ITU-T P.1203.3] Recommendation ITU-T P.1203.3 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport –Quality integration module*.
+- [ITU-T P.1401] Recommendation ITU-T P.1401 (2012), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*.
+
+# 3 Definitions
+
+## 3.1 Terms defined elsewhere
+
+This Recommendation uses the following term defined elsewhere:
+
+- 3.1.1 mean opinion score (MOS):** [ITU-T P.800.1].
+
+Further terms are defined in Recommendation [ITU-T P.1203].
+
+### **3.2 Terms defined in this Recommendation**
+
+None.
+
+## **4 Abbreviations and acronyms**
+
+This Recommendation uses the following abbreviations and acronyms:
+
+|        |                                        |
+|--------|----------------------------------------|
+| AAC    | Advanced Audio Coding                  |
+| AAC-LC | Advanced Audio Coding – Low Complexity |
+| AC3    | Audio Coding 3                         |
+| ACR    | Absolute Category Rating               |
+| ARQ    | Automatic Repeat Request               |
+| FEC    | Forward Error Correction               |
+| HE-AAC | High-Efficiency Advanced Audio Coding  |
+| HTTP   | Hypertext Transfer Protocol            |
+| IP     | Internet Protocol                      |
+| MOS    | Mean Opinion Score                     |
+| MPEG   | Moving Pictures Expert Group           |
+| OTT    | Over The Top                           |
+| TCP    | Transport Control Protocol             |
+| TS     | Transport Stream                       |
+| UDP    | User Datagram Protocol                 |
+
+## **5 Conventions**
+
+None.
+
+# **6 Pa module in ITU-T P.1203 context**
+
+The overall model structure is shown in Figure 6-1, highlighting the position of the *Pa* module. More details on the general structure can be found in the introductory [ITU-T P.1203].
+
+![Block diagram of the ITU-T P.1203 model showing the Pa module in context. The diagram shows a stream I.01 entering 'Media parameter extraction' and 'Buffer parameter extraction' blocks. These lead to 'Input information' (I.11, I.13, I.14). I.11 goes to 'Pa: Audio quality estimation module (ITU-T P.1203.2)'. I.13 goes to 'Pv: Video quality estimation module (ITU-T P.1203.1)'. I.14 goes to 'Pb: quality impact due to buffering'. A box 'I. GEN: Device info available to all modules' feeds into 'Pb'. 'Pa' outputs 0.21 and 0.22. 'Pv' outputs 0.34 and 0.35. 'Pb' outputs 0.23. These are fed into 'Pq: Quality integration module (ITU-T P.1203.3)' and 'Pav: A/V integration/temporal'. The final output is 'Integral MOS' of 0.46. The diagram is labeled P.1203.2(16)_F6-1.](cfef993dcc8fb513de79eb1f93cf26ae_img.jpg)
+
+Block diagram of the ITU-T P.1203 model showing the Pa module in context. The diagram shows a stream I.01 entering 'Media parameter extraction' and 'Buffer parameter extraction' blocks. These lead to 'Input information' (I.11, I.13, I.14). I.11 goes to 'Pa: Audio quality estimation module (ITU-T P.1203.2)'. I.13 goes to 'Pv: Video quality estimation module (ITU-T P.1203.1)'. I.14 goes to 'Pb: quality impact due to buffering'. A box 'I. GEN: Device info available to all modules' feeds into 'Pb'. 'Pa' outputs 0.21 and 0.22. 'Pv' outputs 0.34 and 0.35. 'Pb' outputs 0.23. These are fed into 'Pq: Quality integration module (ITU-T P.1203.3)' and 'Pav: A/V integration/temporal'. The final output is 'Integral MOS' of 0.46. The diagram is labeled P.1203.2(16)\_F6-1.
+
+Figure 6-1 – *Pa* module in context of building blocks of the ITU-T P.1203 model
+
+## 6.1 *Pa* module modes
+
+The modes of operation for ITU-T P.1203.2 are defined in the Table 6-1. Detailed information on exactly which inputs are available for each mode is provided in Table 7-1. A single model is specified for all modes and is described in clause 8.
+
+Table 6-1 –ITU-T P.1203.2 modes
+
+| Mode | Encryption                                      | Input                                               | Complexity                  | Comments                              |
+|------|-------------------------------------------------|-----------------------------------------------------|-----------------------------|---------------------------------------|
+| 0    | Encrypted media payload and media frame headers | Meta-data                                           | Low                         | Module defined in this Recommendation |
+| 1    | Encrypted media payload                         | Meta-data and frame header information              | Low<br>(see comments)       | Same as mode 0                        |
+| 2    | No encryption                                   | Meta-data and up-to 2% of the media stream          | Medium<br>(see comments)    | Same as mode 0                        |
+| 3    | No encryption                                   | Meta-data and any information from the video stream | Unlimited<br>(see comments) | Same as mode 0                        |
+
+## 7 Model input
+
+The model receives media information and prior knowledge about the media stream. The audio quality module receives the following input signals, regardless of the mode of operation, following the measurement window-based procedure as specified in [ITU-T P.1203], clause 7.4:
+
+I.11: Audio coding information, as specified in [ITU-T P.1203], clause 7.1.
+
+Note that fault correction techniques, such as automatic repeat request (ARQ) and forward error correction (FEC) used for user datagram protocol (UDP) based streaming are not applicable for this case, where the streaming is TCP based. In TCP-based transport all retransmissions and packet loss information is typically handled transparently by the transport layer and while it can be available to
+
+the models described in this Recommendation it is not needed. The only information provided to the model that may implicitly include effects such as packet loss and respective retransmission is the initial loading delay and stalling information provided to the quality integration module (see [ITU-T P.1203.3]).
+
+### 7.1 I.11 input specification
+
+Since the audio quality module for mode 0 will be used as a component for the ITU-T P.1203 series models, I.11 consists of two parameters:
+
+- Audio codec
+- Bit rate in kbit/s
+
+Details can be found in Table 7-1.
+
+Note that the actual information available to the module at a specific output sample timestamp is restricted by the measurement window as defined in [ITU-T P.1203], clause 7.4.
+
+**Table 7-1 – Description of I.11**
+
+| ID           | Description                                       | Values                                    | Frequency         | Modes available |
+|--------------|---------------------------------------------------|-------------------------------------------|-------------------|-----------------|
+| <b>I.GEN</b> |                                                   |                                           |                   |                 |
+| 0            | The resolution of the image displayed to the user | Number of pixels (WxH) in displayed video | Per media session | All             |
+| 1            | The device type on which the media is played      | PC or mobile                              | Per media session | All             |
+| <b>I.11</b>  |                                                   |                                           |                   |                 |
+| 2            | Target audio bit rate                             | Bit rate in kbit/s.                       | Per media segment | All             |
+| 3            | Segment duration                                  | Duration in seconds                       | Per media segment | All             |
+| 4            | Audio frame number                                | Integer, starting with 1                  | Per media segment | 1,2,3           |
+| 5            | Audio frame size                                  | Size of the frame in bytes                | Per audio frame   | 1,2,3           |
+| 6            | Audio frame duration                              | Duration in seconds                       | Per audio frame   | 1,2,3           |
+| 7            | Audio codec                                       | One of: AAC-LC, AAC-HEv1, AAC-HEv2, AC3   | Per media segment | All             |
+| 8            | Audio sampling frequency                          | Hz                                        | Per media segment | All             |
+| 9            | Number of audio channels                          | 2                                         | Per media segment | All             |
+| 10           | Audio bit-stream                                  | Encoded audio bytes for the frame         | Per audio frame   | 2,3             |
+
+## 8 Model algorithm and output
+
+The [ITU-T P.1203.2] model for audio has one output, O.21. It provides output values on the 5-point ACR scale ("MOS") per output sampling interval.
+
+One single audio quality module is recommended to be used in the ITU-T P.1203 series models. This audio quality module algorithm is the same as the one specified in [ITU-T P.1201.2]. It is summarized here for completeness:
+
+$$O.21 = MOSfromR(QA) \quad (\text{Eq. 13d in [ITU-T P.1201.2]})$$
+
+with:
+
+$$QA = 100 - QcodA \quad (\text{Eq. 13c in [ITU-T P.1201.2]})$$
+
+with coding degradations only, i.e., with  $QtraA = 0$ )
+
+where:
+
+$$QcodA = a1A \times \exp(a2A \times Bitrate) + a3A \quad (\text{Eq. 13a in [ITU-T P.1201.2]})$$
+
+Bit rate is the audio bit rate in kBit/s.
+
+The function  $MOSfromR$  is given in Annex E of [ITU-T P.1203.1] and is provided below:
+
+$$MOSfromR: \mathbb{R} \mapsto \mathbb{R}$$
+
+$$Q \mapsto MOS := MOSfromR(Q)$$
+
+$$MOS = MOS_{MIN} + (MOS_{MAX} - MOS_{MIN}) * \frac{Q}{100} + Q * (Q - 60) * (100 - Q) * 0.000007 \quad (E.1)$$
+
+$$MOS = \min(MOS_{MAX}, \max(MOS, MOS_{MIN})) \quad (E.2)$$
+
+where  $MOS_{MAX} = 4.9$  and  $MOS_{MIN} = 1.05$ .
+
+Coefficients  $a1A$ ,  $a2A$  and  $a3A$  depend on the audio codec. These audio model coefficients are provided in Table 8-1:
+
+**Table 8-1 – Audio model coefficients for different audio codecs (coding degradations only), adapted from Table 1 of [ITU-T P.1201.2]**
+
+| Audio codec | a1A   | a2A   | a3A   |
+|-------------|-------|-------|-------|
+| MPEG1 L2    | 100.0 | -0.02 | 15.48 |
+| AC3         | 100.0 | -0.03 | 15.70 |
+| AAC-LC      | 100.0 | -0.05 | 14.60 |
+| HE-AAC v2   | 100.0 | -0.11 | 20.06 |
+
+## Bibliography
+
+- [b-Zielinski\_2008] Slawomir Zielinski, Soren Bech and Francis Rumsey (2008), *On some biases encountered in modern audio quality listening tests – A review*, *Journal Audio Engineering Society (JAES)*, 56(6), 427-451.  
+<<http://www.aes.org/e-lib/browse.cfm?elib=14393>>
+
+
+
+## SERIES OF ITU-T RECOMMENDATIONS
+
+|                 |                                                                                                                                                           |
+|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------|
+| Series A        | Organization of the work of ITU-T                                                                                                                         |
+| Series D        | Tariff and accounting principles and international telecommunication/ICT economic and policy issues                                                       |
+| Series E        | Overall network operation, telephone service, service operation and human factors                                                                         |
+| Series F        | Non-telephone telecommunication services                                                                                                                  |
+| Series G        | Transmission systems and media, digital systems and networks                                                                                              |
+| Series H        | Audiovisual and multimedia systems                                                                                                                        |
+| Series I        | Integrated services digital network                                                                                                                       |
+| Series J        | Cable networks and transmission of television, sound programme and other multimedia signals                                                               |
+| Series K        | Protection against interference                                                                                                                           |
+| Series L        | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant |
+| Series M        | Telecommunication management, including TMN and network maintenance                                                                                       |
+| Series N        | Maintenance: international sound programme and television transmission circuits                                                                           |
+| Series O        | Specifications of measuring equipment                                                                                                                     |
+| <b>Series P</b> | <b>Telephone transmission quality, telephone installations, local line networks</b>                                                                       |
+| Series Q        | Switching and signalling, and associated measurements and tests                                                                                           |
+| Series R        | Telegraph transmission                                                                                                                                    |
+| Series S        | Telegraph services terminal equipment                                                                                                                     |
+| Series T        | Terminals for telematic services                                                                                                                          |
+| Series U        | Telegraph switching                                                                                                                                       |
+| Series V        | Data communication over the telephone network                                                                                                             |
+| Series X        | Data networks, open system communications and security                                                                                                    |
+| Series Y        | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities                               |
+| Series Z        | Languages and general software aspects for telecommunication systems                                                                                      |

marked/P/T-REC-P.1204.3-202001-I_PDF-E/raw.md

@@ -0,0 +1,770 @@
+
+
+**ITU-T**
+
+TELECOMMUNICATION  
+STANDARDIZATION SECTOR  
+OF ITU
+
+**P.1204.3**
+
+(01/2020)
+
+SERIES P: TELEPHONE TRANSMISSION QUALITY,  
+TELEPHONE INSTALLATIONS, LOCAL LINE  
+NETWORKS
+
+Models and tools for quality assessment of streamed  
+media
+
+---
+
+**Video quality assessment of streaming services  
+over reliable transport for resolutions up to 4K  
+with access to full bitstream information**
+
+Recommendation ITU-T P.1204.3
+
+# ITU-T P-SERIES RECOMMENDATIONS
+
+## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS
+
+|                                                                                                    |                      |
+|----------------------------------------------------------------------------------------------------|----------------------|
+| Vocabulary and effects of transmission parameters on customer opinion of transmission quality      | P.10–P.19            |
+| Voice terminal characteristics                                                                     | P.30–P.39            |
+| Reference systems                                                                                  | P.40–P.49            |
+| Objective measuring apparatus                                                                      | P.50–P.59            |
+| Objective electro-acoustical measurements                                                          | P.60–P.69            |
+| Measurements related to speech loudness                                                            | P.70–P.79            |
+| Methods for objective and subjective assessment of speech quality                                  | P.80–P.89            |
+| Voice terminal characteristics                                                                     | P.300–P.399          |
+| Objective measuring apparatus                                                                      | P.500–P.599          |
+| Measurements related to speech loudness                                                            | P.700–P.709          |
+| Methods for objective and subjective assessment of speech and video quality                        | P.800–P.899          |
+| Audiovisual quality in multimedia services                                                         | P.900–P.999          |
+| Transmission performance and QoS aspects of IP end-points                                          | P.1000–P.1099        |
+| Communications involving vehicles                                                                  | P.1100–P.1199        |
+| <b>Models and tools for quality assessment of streamed media</b>                                   | <b>P.1200–P.1299</b> |
+| Telemeeting assessment                                                                             | P.1300–P.1399        |
+| Statistical analysis, evaluation and reporting guidelines of quality measurements                  | P.1400–P.1499        |
+| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500–P.1599        |
+
+For further details, please refer to the list of ITU-T Recommendations.
+
+# Recommendation ITU-T P.1204.3
+
+# Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full bitstream information
+
+## Summary
+
+Recommendation ITU-T P.1204.3 describes a bitstream-based mode 3 video quality model for monitoring the video quality for streaming using reliable transport (e.g., hypertext transfer protocol- (HTTP-) based adaptive streaming (HAS) over the transmission control protocol (TCP), quick user datagram protocol internet connections (QUIC)). The estimate is validated for videos encoded with H.264, H.265 or video payload type 9 (VP9) codecs at any resolution up to 4K/ultra-high definition (UHD) resolution for personal computer (PC) monitors and television (TV) and up to  $2\,560 \times 1\,440$  for smartphone and tablet displays.
+
+The ITU-T P.1204 series of Recommendations provide sequence-related (between 5 s and 10 s) and per-1-second video-quality estimation. In principle, the per-one-second outputs of this video-quality model can be used together with an audio model for integration into audiovisual quality and, together with information about initial loading delay and media playout stalling events, further into a final per-session model output, an estimate of integral per-session quality (see e.g., ITU-T P.1203, ITU-T P.1203.2, ITU-T P.1203.3).
+
+Recommendation ITU-T P.1204.3 was developed in collaboration with the Video Quality Experts Group (VQEG).
+
+The ITU-T P.1204 series of Recommendations addresses three application areas:
+
+- large-screen presentation as with fixed-network video streaming;
+- mobile streaming on handheld devices such as smartphones;
+- presentation on tablet-type devices.
+
+This Recommendation includes an electronic attachment with the Trees for final prediction announced in clause 8.2.
+
+## History
+
+| Edition | Recommendation | Approval   | Study Group | Unique ID*                                                                |
+|---------|----------------|------------|-------------|---------------------------------------------------------------------------|
+| 1.0     | ITU-T P.1204.3 | 2020-01-13 | 12          | <a href="http://handle.itu.int/11.1002/1000/14156">11.1002/1000/14156</a> |
+
+## Keywords
+
+Adaptive streaming, IPTV, mean opinion score (MOS), mobile video, mobile TV, monitoring, multimedia, OTT, progressive download, QoE, TV, video.
+
+---
+
+\* To access the Recommendation, type the URL <http://handle.itu.int/> in the address field of your web browser, followed by the Recommendation's unique ID. For example, <http://handle.itu.int/11.1002/1000/11830-en>.
+
+## FOREWORD
+
+The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis.
+
+The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics.
+
+The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1.
+
+In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC.
+
+## NOTE
+
+In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency.
+
+Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party.
+
+## INTELLECTUAL PROPERTY RIGHTS
+
+ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process.
+
+As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at <http://www.itu.int/ITU-T/ipr/>.
+
+© ITU 2020
+
+All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU.
+
+## Table of Contents
+
+|                                                                 | Page |
+|-----------------------------------------------------------------|------|
+| 1 Scope.....                                                    | 1    |
+| 2 References.....                                               | 2    |
+| 3 Definitions .....                                             | 3    |
+| 3.1 Terms defined elsewhere .....                               | 3    |
+| 3.2 Terms defined in this Recommendation.....                   | 3    |
+| 4 Abbreviations and acronyms .....                              | 3    |
+| 5 Conventions .....                                             | 4    |
+| 6 Areas of application.....                                     | 4    |
+| 6.1 Application range for the model.....                        | 4    |
+| 7 Model algorithm and output .....                              | 6    |
+| 7.1 Building blocks in relation ITU-T P.1204 model context..... | 6    |
+| 7.2 Model input interfaces .....                                | 7    |
+| 7.3 Specification of inputs I.GEN, I.13 .....                   | 7    |
+| 7.4 Model output information.....                               | 8    |
+| 8 Model architecture of this Recommendation.....                | 8    |
+| 8.1 Parametric part – The core model .....                      | 9    |
+| 8.2 Machine-learning-based part of the model.....               | 11   |
+| 8.3 Final prediction.....                                       | 12   |
+| 8.4 Per-second score prediction.....                            | 13   |
+| Annex A – Helper function definitions.....                      | 14   |
+| Appendix I – Performance figures .....                          | 16   |
+| Bibliography.....                                               | 17   |
+
+
+
+# Recommendation ITU-T P.1204.3
+
+# Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full bitstream information
+
+# 1 Scope
+
+This Recommendation<sup>1</sup> describes a bitstream-based video quality model that can be used: stand-alone as a video quality prediction model; or together with audio and integration modules to form a complete model to predict the impact of audio and video media encodings and observed Internet protocol (IP) network impairments on quality experienced by the end-user in multimedia streaming applications. The streaming techniques addressed comprise progressive download and adaptive streaming, for both mobile and fixed network streaming applications.
+
+This model is defined to cover a range of use cases, from monitoring bitstreams where the video payload is fully encrypted, unencrypted bitstreams and where deep packet inspection is possible or where the bitstream is available at the encoding premises, e.g., from the client side. The model thus has a wide range of application, from encoding optimization over client-side quality of experience (QoE) assessment to network or service optimization or benchmarking purposes. The model in this Recommendation is bitstream based.
+
+The model described here is applicable to progressive download and adaptive streaming or other streaming applications with reliable transport, where the quality experienced by the end user is affected by video degradations due to coding, spatial re-scaling or variations in video frame rates. Quality assessment of adaptive streaming includes aspects of media adaptation that may be handled in integration modules such as those of [ITU-T P.1203.3] and not in the video modules in this Recommendation. This Recommendation is able to handle various video codecs (i.e., H.264, H.265/high-efficiency video coding (HEVC) and video payload type 9 (VP9), resolutions up to 4K/ultra-high definition-1 (UHD-1) and frame rates up to 60 frames/s. In contrast to the video-quality module Pv of [b-ITU-T P.1203], i.e., [ITU-T P.1203.1], only addresses ITU-T H.264 and full high definition (HD) with up to 30 frames/s.
+
+The model predicts a mean opinion score (MOS) on a five-point absolute category rating (ACR) scale (see [ITU-T P.910]) as an overall video quality MOS (5 s to 10 s). In addition to the overall quality score, this video quality model produces a per-one-second quality score, suitable for diagnostics or integration into an integral quality score for longer sessions (see, for example [ITU-T P.1203.3] for 1 min to 5 min duration sessions).
+
+The model associated with this Recommendation cannot provide a comprehensive evaluation of the video quality as perceived by an *individual end-user* because the scores reflect the perceived impairments due to coded video media data being transmitted over an IP connection with certain performance and do not include specific terminal device or user-specific information. The scores predicted by such a general quality model necessarily reflect *average perceptual quality*.
+
+Effects due to source generations, such as signal noise, video shake, certain colour properties (and other similar video factors) and other impairments related to the payload, are not reflected in the scores computed by this model.
+
+As a consequence, this Recommendation can be used for applications such as:
+
+- in-service quality monitoring for specific IP-based audiovisual services, as specified in more detail in clause 6.1;
+
+---
+
+<sup>1</sup> This Recommendation includes an electronic attachment with the Trees for final prediction announced in clause 8.2.
+
+- performance and quality assessment of live networks (including codecs) considering the effect due to encoding bitrate, encoding resolution and encoding frame rate;
+- laboratory testing of video systems;
+- benchmarking of different service implementations.
+
+In particular, targeted applications are progressive download streaming and adaptive streaming (using reliable transport), which includes the following.
+
+- Over-the-top (OTT) services, as well as operator-managed video services (over the TCP).
+- Video over both mobile and fixed connections.
+- The streaming protocols HTTP live streaming (HLS) or dynamic adaptive streaming over HTTP (DASH) used with the hypertext transfer protocol (HTTP) or HTTP2 over TCP/IP or quick user datagram protocol internet connections (QUIC), or real-time messaging protocol (RTMP) over TCP/IP. Note that the model is agnostic to the specific application or transport layer protocol, with the exception that it assumes reliable delivery of video packets.
+- Video services typically using container formats based on the ISO/IEC base media file format such as Moving Picture Experts Group-4 (MPEG-4) Part 14 (MP4), or other container formats such as audio video interleave (AVI), Matroska video (MKV), WebM, Third Generation Partnership (3GP), and MPEG-2 transport stream (MPEG2-TS). Note that the model is agnostic to the type of container format.
+
+# 2 References
+
+The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation.
+
+- [ITU-T H.264] Recommendation ITU-T H.264 (2019), *Advanced video coding for generic audiovisual services*.
+- [ITU-T H.265] Recommendation ITU-T H.265 (2019), *High efficiency video coding*.
+- [ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*.
+- ITU-T P.1203.1] Recommendation ITU-T P.1203.1 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Video quality estimation module*.
+- [ITU-T P.1203.3] Recommendation ITU-T P.1203.3 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Quality integration module*.
+- [ITU-T P.1204] Recommendation ITU-T P.1204 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K*.
+
+# 3 Definitions
+
+## 3.1 Terms defined elsewhere
+
+This Recommendation uses the following term defined elsewhere:
+
+**3.1.1 bitstream** [ITU-T H.264]: A sequence of bits that forms the representation of coded pictures and associated data forming one or more coded video sequences. Bitstream is a collective term used to refer either to a NAL unit stream or a byte stream.
+
+**3.1.2 mean opinion score (MOS)** [ITU-T P.1204]: The mean of opinion scores, which are values on a predefined scale that subjects assign to their opinion of the performance of the telephone transmission system used either for conversation or for listening to spoken material.
+
+NOTE – Paraphrased from clause 7 of [b-ITU-T P.800.1].
+
+**3.1.3 media adaptation** [b-ITU-T P.1203]: Events where the player switches video playback between a known set of media quality levels while adapting to network conditions, by downloading and decoding individual segments in sequence.
+
+**3.1.4 integral quality** [b-ITU-T P.1203]: The quality as perceived by a subject in a subjective test, which corresponds to the scope of this Recommendation. Artefacts presented in the subjective tests typically include a combination of audio compression, video compression, and stalling effects.
+
+**3.1.5 media quality level** [b-ITU-T P.1203]: A particular encoding setting applied to a video or audio stream.
+
+**3.1.6 model, model algorithm** [b-ITU-T P.1203]: An algorithm with the purpose of estimating the subjective (perceived) quality of a media sequence.
+
+**3.1.7 sequence** [b-ITU-T P.1203]: An audiovisual stream composed of multiple non-overlapping segments.
+
+**3.1.8 video chunk** [b-ITU-T G.1022]: A contiguous set of samples for one track of a video.
+
+## 3.2 Terms defined in this Recommendation
+
+None.
+
+# 4 Abbreviations and acronyms
+
+This Recommendation uses the following abbreviations and acronyms:
+
+|      |                                      |
+|------|--------------------------------------|
+| ACR  | Absolute Category Rating             |
+| AV1  | AOMedia Video 1                      |
+| AVI  | Audio Video Interleave               |
+| DASH | Dynamic Adaptive Streaming over HTTP |
+| GoP  | Group of Pictures                    |
+| HAS  | HTTP-based adaptive streaming        |
+| HD   | High Definition                      |
+| HEVC | High-Efficiency Video Coding         |
+| HLS  | HTTP Live Streaming                  |
+| HTTP | Hypertext Transfer Protocol          |
+| I-   | Intra-predicted                      |
+| IP   | Internet Protocol                    |
+
+|           |                                                   |
+|-----------|---------------------------------------------------|
+| IQR       | Interquartile Range                               |
+| MKV       | Matroska Video                                    |
+| MOS       | Mean Opinion Score                                |
+| MP4       | MPEG-4 Part 14                                    |
+| MPEG      | Moving Pictures Expert Group                      |
+| MPEG-2-TS | MPEG-2 Transport Stream                           |
+| OTT       | Over The Top                                      |
+| PC        | Personal Computer                                 |
+| QHD       | Quad High Definition                              |
+| QoE       | Quality of Experience                             |
+| QUIC      | Quick User datagram protocol Internet Connections |
+| Rext      | Range extension                                   |
+| RMSE      | Root Mean Square Error                            |
+| RTMP      | Real-Time Messaging Protocol                      |
+| RTP       | Real-time Transport Protocol                      |
+| TCP       | Transmission Control Protocol                     |
+| TV        | Television                                        |
+| UDP       | User Datagram Protocol                            |
+| UHD       | Ultra-High Definition                             |
+| VP9       | Video Payload type 9                              |
+| VVC       | Versatile Video Coding                            |
+
+# 5 Conventions
+
+This Recommendation uses the following conventions:
+
+- 4K: Video resolution of  $4\,096 \times 2\,160$  or  $3\,840 \times 2\,160$ ;
+- Pv designates the video quality estimation module (as specified in this Recommendation for the case of bitstream-based prediction, see [ITU-T P.1204] for alternative implementations such as pixel based and hybrid);
+- Reliable transport: Reliable delivery with protocols guaranteeing no loss of information.
+
+# 6 Areas of application
+
+## 6.1 Application range for the model
+
+Table 1 shows the application range of the model in this Recommendation based on what the model has actually been developed for and Table 2 lists areas where it is not applicable. Table 3 lists test factors and coding technologies for which this Recommendation has been validated.
+
+**Table 1 – Areas for which this Recommendation is applicable**
+
+| <b>Areas for which the model is applicable</b>                                                                                                                                                                                                                                               |
+|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| In-service monitoring of video sent over reliable transport. Both OTT services and operator-managed video services, using reliable delivery with protocols such as HTTP or HTTP2 over TCP/IP or QUIC, or RTMP over TCP/IP. Note that this model is agnostic to the type of container format. |
+| Performance and quality assessment of live networks (including video encoding) considering impairments due to encoding bitrate, encoding resolution, and encoding frame rate.                                                                                                                |
+| Laboratory testing of video systems.                                                                                                                                                                                                                                                         |
+| Benchmarking of different service implementations.                                                                                                                                                                                                                                           |
+
+**Table 2 – Areas for which this Recommendation is not applicable**
+
+| <b>Areas for which the model is not applicable</b>                                                                                                                                                   |
+|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| In-service monitoring of video streaming using unreliable transport (e.g., real-time transport protocol/user datagram protocol (RTP/UDP)), where packet loss introduces visible quality degradations |
+| Evaluation of visual quality of display/device properties                                                                                                                                            |
+| Evaluation of audio/video sync distortions                                                                                                                                                           |
+| Evaluation of video codecs for which the model is not validated (AOMedia Video 1(AV1), MPEG-I Part 3 [versatile video coding (VVC)], etc.)                                                           |
+| Evaluation of the effects of noise, delay, colour correctness or other content-production-related aspects                                                                                            |
+
+**Table 3 – Test factors, and coding technologies for which this Recommendation has been validated**
+
+| <b>Video test factors for which the model has been validated</b> |                                                                                                                                                                                                                                                                                       |                           |                                              |                              |
+|------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------|----------------------------------------------|------------------------------|
+| Video content                                                    | Movies and movie trailers, sports videos, documentaries, computer generated graphics/games, etc.                                                                                                                                                                                      |                           |                                              |                              |
+| Input video length                                               | The video modules were trained and validated to produce one overall video-quality score for a chunk of ~7–9 s and also provide the per-second scores. Optimal performance for ~ 8 s. Models are assumed to provide valid overall video-quality estimations for 5–10 s long sequences. |                           |                                              |                              |
+| Bitstream Container                                              | AVI, MP4, MKV, WebM                                                                                                                                                                                                                                                                   |                           |                                              |                              |
+| Encoder types (and implementation, see Note 1)                   | H.264/AVC (libx264), H.265/HEVC (libx265), VP9 (libvpx-vp9)                                                                                                                                                                                                                           |                           |                                              |                              |
+| Encoder profiles                                                 | H.264 (MPEG-4 Part 10): Constrained baseline, Main, Hi, Hi10, Hi422.<br>H.265: Main, Main10, range extension (Rext).<br>VP9: 0, 1, 2, 3.                                                                                                                                              |                           |                                              |                              |
+| Video resolution and bitrate                                     | <b>Resolution definition</b>                                                                                                                                                                                                                                                          | <b>Video height range</b> | <b>Personal computer/ television (PC/TV)</b> | <b>Mobile/tablet (MO/TA)</b> |
+|                                                                  | Below SD                                                                                                                                                                                                                                                                              | 180-270                   | —                                            | 90 Kbps-1 Mbps               |
+|                                                                  | SD                                                                                                                                                                                                                                                                                    | 360-540                   | 150 Kbps-4 Mbps                              | 150 Kbps-4 Mbps              |
+|                                                                  | HD                                                                                                                                                                                                                                                                                    | 720-1 080                 | 500 Kbps-15 Mbps                             | 500 Kbps-15 Mbps             |
+|                                                                  | Above HD                                                                                                                                                                                                                                                                              | 1 440-2 160               | 1.5 Mbps-45 Mbps                             | 1.5 Mbps-20 Mbps             |
+
+**Table 3 – Test factors, and coding technologies for which this Recommendation  
+has been validated**
+
+| <b>Video test factors for which the model has been validated</b>                                                                                                                                                                                                                                                                                                                                                                                                                                                                     |                                                                                                                                                  |
+|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------|
+| Video aspect ratio                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   | 16:9, see Note 2                                                                                                                                 |
+| Group of pictures (GoP)                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              | Variable. Average GOP length can be between 0.5 s and chunk duration                                                                             |
+| Bit-depth                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            | 8 bit or 10 bit                                                                                                                                  |
+| Chroma subsampling                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   | YUV 4:2:0 and YUV 4:2:2                                                                                                                          |
+| OTTs                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 | Online providers that offer video on demand and video encoding as a service. It should be noted that the models are applicable for similar OTTs. |
+| Display resolution and frame rate                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    | PC/TV; 2 160 p, up to 60 frames/s.<br>MO/TA: 1 440 p, up to 60 frames/s.                                                                         |
+| Viewing distances                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    | PC/TV: 1.5 <i>H</i> to 3 <i>H</i> ( <i>H</i> : Screen height), see Note 3<br>MO/TA: 4 <i>H</i> to 6 <i>H</i>                                     |
+| NOTE 1 – During training and validation, FFmpeg 3.2.2 was used with x264 snapshot 20170202-2245, x265 v2.2, libvpx 1.6.1.<br>NOTE 2 – For original content with a larger aspect ratio, letterboxing of up to 30% was allowed, that is 1 512 pixels height for video coded at 2 160 pixels height. Video content with 1.89:1 aspect ratio (e.g., cinema 4K) may also be used.<br>NOTE 3 – It is noted that for PC/MO, the model output is conservative and should be interpreted to correspond to a viewing distance of 1.5H to 1.6H. |                                                                                                                                                  |
+
+**7 Model algorithm and output**
+
+**7.1 Building blocks in relation ITU-T P.1204 model context**
+
+The module layout of the ITU-T P.1204 model is depicted in Figure 1.
+
+![Figure 1 – Building blocks of the bitstream-based video quality model of this Recommendation (PvP.1204.3) and input information processing](602ada2a012ff3cc38d91de2eec5b450_img.jpg)
+
+```
+
+  graph LR
+    Stream[Stream I.01] --> Decoder[Pixel information extraction decoder]
+    Stream --> Media[Media parameter extraction]
+    Media -- I.13 --> Bitstream[Bitstream-based ITU-T P.1204.3]
+    Decoder --> Bitstream
+    Input[Input information] -.-> Bitstream
+    Device[I.GEN: Device info available to all modules] --> Bitstream
+    Bitstream -- Pv ITU-T P.1204 --> O27[O.27 Video quality MOS, 5-10 sec]
+    Bitstream --> O22[O.22 5-point per-1-sec video quality]
+  
+```
+
+Figure 1 – Building blocks of the bitstream-based video quality model of this Recommendation (PvP.1204.3) and input information processing
+
+**Figure 1 – Building blocks of the bitstream-based video quality model of this  
+Recommendation (Pv<sub>P.1204.3</sub>) and input information processing**
+
+6 Rec. ITU-T P.1204.3 (01/2020)
+
+## 7.2 Model input interfaces
+
+The model receives the following input information:
+
+**I.GEN:** Display resolution and device type. The device type is defined as follows:
+
+- PC/TV: screen size 24 inch or larger and less than or equal to 100 inches.
+- MO/TA: screen size 13 inch or smaller.
+
+**I.13:** Video coding information
+
+## 7.3 Specification of inputs I.GEN, I.13
+
+See Table 4.
+
+**Table 4 – I.GEN and I.13 inputs description (see Note 1)**
+
+| ID                  | Description                                             | Values                                                                                                                       | Frequency       | Used in this Recommendation |
+|---------------------|---------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------|-----------------|-----------------------------|
+| <b><i>I.GEN</i></b> |                                                         |                                                                                                                              |                 |                             |
+| 0                   | The resolution of the image displayed to the user       | Number of pixels ( $W \times H$ ) in displayed video                                                                         | Per media chunk | Yes                         |
+| 1                   | The device type on which the media is played            | "PC", "TV", "MO", "TA"                                                                                                       | Per media chunk | Yes                         |
+| 2                   | Device display size                                     | Display size (diagonal in inches)                                                                                            | Per media chunk | Yes                         |
+| 3                   | Relative viewing distance in multiple of display height | Relative viewing distance                                                                                                    | Per media chunk | Yes                         |
+| <b><i>I.13</i></b>  |                                                         |                                                                                                                              |                 |                             |
+| 4                   | Video bitrate                                           | Bitrate in kilobits per second                                                                                               | Per media chunk | Yes                         |
+| 5                   | Video frame rate                                        | Frame rate in frames per second.                                                                                             | Per media chunk | Yes                         |
+| 6                   | Segment duration                                        | Duration in seconds                                                                                                          | Per media chunk | Yes                         |
+| 7                   | Video encoding resolution                               | Number of pixels ( $W \times H$ ) in transmitted video                                                                       | Per media chunk | Yes                         |
+| 8                   | Video codec and profile                                 | H.264 (MPEG-4 Part 10):<br>Constrained Baseline, Main, Hi,<br>Hi10, Hi422.<br>H.265: Main, Main10, Rext.<br>VP9: 0, 1, 2, 3. | Per media chunk | Yes                         |
+| 9                   | Video frame number                                      | Integer, starting at 1, denoting the frame sequence number in encoding order                                                 | Per video frame | No                          |
+| 10                  | Video frame duration                                    | Duration of the frame in seconds                                                                                             | Per video frame | Yes                         |
+| 13                  | Video frame size                                        | The size of the encoded video frame in bytes                                                                                 | Per video frame | Yes                         |
+| 14                  | Type of each picture                                    | See Note 2.<br>"I"/"P"/"B" for this Recommendation                                                                           | Per video frame | Yes                         |
+
+**Table 4 – I.GEN and I.13 inputs description (see Note 1)**
+
+| ID | Description        | Values                                                            | Frequency       | Used in this Recommendation |
+|----|--------------------|-------------------------------------------------------------------|-----------------|-----------------------------|
+| 15 | Video bitstream    | Encoded video bytes for the frame                                 | Per video frame | Yes                         |
+| 16 | Video pixel format | 8-bit or 10-bit, together with 4:2:2 or 4:2:0 chroma subsampling. | Per media chunk | Yes                         |
+
+NOTE 1 – This table will also address ITU-T P.1204.1 and ITU-T P.1204.2 once these have been approved by ITU-T.  
+NOTE 2 – Other values are under study.
+
+## 7.4 Model output information
+
+The video module defined in this Recommendation had two outputs, O.22 and O.27. It provides output values on the five-point ACR scale (MOS).
+
+# 8 Model architecture of this Recommendation
+
+The general model structure is shown in Figure 2. The model consists of two parts, namely, parametric and machine-learning parts. The machine-learning part of the model is based on random forests. The overall prediction is a weighted sum of the parametric and machine-learning part predictions.
+
+![Figure 2 – General model structure. The diagram shows a flow from 'Video segment' to 'Bitstream parser' to 'Feature aggregation' (e.g., per GoP, per segment, etc.). This leads to a box containing 'VP9', 'H264', 'H265', 'Coding degradation', 'Upscaling degradation', and 'Framerate degradation'. This box feeds into 'Initial parametric prediction'. Below this, a 'Random forest for delta prediction' feeds into 'Delta prediction'. The 'Initial parametric prediction' and 'Delta prediction' are combined via a summation node (+) with weights w1 and w2 respectively, leading to the 'Final prediction'.](1b5a812c8aa20fd5cba28e97001d32de_img.jpg)
+
+```
+
+graph LR
+    VS[Video segment] --> BP[Bitstream parser]
+    BP --> FA[Feature aggregation  
+e.g., per GoP,  
+per segment,  
+etc.]
+    FA --> Degradations["VP9  
+H264  
+H265  
+Coding degradation  
+Upscaling degradation  
+Framerate degradation"]
+    Degradations --> IPP[Initial parametric prediction]
+    RFD[Random forest for delta prediction] --> DP[Delta prediction]
+    IPP -- w1 --> Sum1((+))
+    DP -- w2 --> Sum1
+    Sum1 --> FP[Final prediction]
+  
+```
+
+P.1204.3(20)\_F02
+
+Figure 2 – General model structure. The diagram shows a flow from 'Video segment' to 'Bitstream parser' to 'Feature aggregation' (e.g., per GoP, per segment, etc.). This leads to a box containing 'VP9', 'H264', 'H265', 'Coding degradation', 'Upscaling degradation', and 'Framerate degradation'. This box feeds into 'Initial parametric prediction'. Below this, a 'Random forest for delta prediction' feeds into 'Delta prediction'. The 'Initial parametric prediction' and 'Delta prediction' are combined via a summation node (+) with weights w1 and w2 respectively, leading to the 'Final prediction'.
+
+**Figure 2 – General model structure**
+
+The model has one output with values on the five-point ACR scale (MOS). The parametric part of the algorithm is the core model. The parametric part of the model  $M_{\text{parametric}}$  is based on the principle of degradation-based modeling. In the proposed approach, three different degradations are identified that may affect the perceived quality of a given video. The general concept is that the higher the degradation, the lower the quality of the video.
+
+The three degradations that affect that quality of a given video are as follows.
+
+- Quantization degradation. This relates to the coding-related degradations that are introduced in videos based on the quantization settings selected. This degradation can be perceived by the end-user as blockiness and other artefacts. The types of artefact and their strength are
+
+codec dependent, as different codecs introduce different distortions based on the selected quantization settings.
+
+- Upscaling degradation. This relates to the degradation introduced due mainly to the encoded video being upscaled to the higher display resolution during playback, thereby resulting in blurring artefacts. These are the same for all codecs, as the display resolution is the only influencing factor for this degradation. It is further assumed that the upscaling algorithm is constant and independent of the codec used, which is the case in real world streaming, where upscaling is performed by the player software or display device used.
+- Temporal degradation: This relates to the degradation introduced due to playing out the distorted video at a reduced frame rate compared to the display's native frame rate, thereby resulting in jerkiness. This is the same for all codecs, as the video frame rate is the only influencing factor for this degradation.
+
+Of the three degradations, only quantization degradation is codec dependent.
+
+## 8.1 Parametric part – The core model
+
+Determination of the quantization degradation:
+
+$$quant = \frac{QP_{\text{non-I-frames}}}{QP_{\text{max}}} \quad (1)$$
+
+where
+
+$QP_{\text{non-I-frames}}$  is the average of the  $QP$  for other than intra-predicted (I-) frames for an entire segment;
+
+$QP_{\text{max}}$  is codec and bit-depth dependent:
+
+- for H.264/H.265 8 Bit  $QP_{\text{max}} = 51$ ,
+- for H.264/H.265 10 Bit  $QP_{\text{max}} = 63$ ,
+- for VP9 8 or 10 Bit  $QP_{\text{max}} = 255$ ;
+
+$quant \in [0, 1]$ .
+
+$$mos_q = a + b * \exp(c * quant + d) \quad (2)$$
+
+$$D_{q\_raw} = 100 - RfromMOS(mos_q) \quad (3)$$
+
+$$D_q = \max(\min(D_{q\_raw}, 100), 0) \quad (4)$$
+
+where  $RfromMOS$  is defined in Annex A.
+
+NOTE – The  $RfromMOS$  and  $MOSfromR$  computations involve information loss due to the fact that these two functions assume that the highest MOS that can be reached is 4.5, thereby resulting in clipping on the MOS-scale for ratings higher than 5. To avoid this information loss, all the subjective data used to train the model is compressed to the 4.5-scale by a simple linear transformation, and the model is trained on this data. Therefore, the resulting coefficients predict the initial prediction on a 4.5-scale. To obtain the prediction on the original five-point scale, the initial prediction is scaled back to the five-scale using the inverse linear transformation.
+
+Determination of the upscaling degradation:
+
+$$scale\_factor = \frac{coding\_res}{display\_res} \quad (5)$$
+
+where
+
+$display\_res = (3840 * 2160)$  for PC/TV and  $(2560 * 1440)$  for MO/TA;
+
+*coding\_res* is the resolution at which the video is encoded (*height \* width*);  
+*scale\_factor*  $\in [0, 1]$ .
+
+$$D_{u\_raw} = x * \log(y * scale\_factor) \quad (6)$$
+
+$$D_u = \max(\min(D_{u\_raw}, 100), 0) \quad (7)$$
+
+Determination of the frame rate degradation:
+
+$$D_{t\_raw} = z * \ln(k * (framerate\_scale\_factor)) \quad (8)$$
+
+where
+
+$$framerate\_scale\_factor = \frac{coding\_framerate}{60}$$
+
+$$framerate\_scale\_factor \in [0, 1]$$
+
+$$D_t = \max(\min(D_{t\_raw}, 100), 0) \quad (9)$$
+
+Parametric part related final MOS:
+
+$$M_{\text{parametric}} = 100 - (D_q + D_u + D_t) \quad (10)$$
+
+$$M_{\text{parametric}} = MOSfromR(M_{\text{parametric}}) \quad (11)$$
+
+$$M_{\text{parametric}} = scaleto5(M_{\text{parametric}}) \quad (12)$$
+
+where *scaleto5* is defined in Annex A.
+
+Scaling is done as the coefficients are trained by compressing the subjective scores to a scale of 4.5 to avoid the information loss that can be introduced by the *RfromMOS* and *MOSfromR* calculations, as noted in this subclause.
+
+### 8.1.1 Model coefficients
+
+The model has access to the entire bitstream as input.
+
+- Coding degradation. It is codec- and bit-depth-dependent. This results in five sets of coefficients, one each for H.264-8bit, H.264-10bit, H.265-8bit, H.265-10bit and VP9 codecs.
+- $QP_{\max}$  is 51 for H.264-8bit, H.265-8bit; 63 for H.264-10bit, H.265-10bit; 255 for VP9.
+
+The model coefficients are listed in Tables 5, 6, 7 and 8.
+
+**Table 5 – Mode 3 – PC/TV**
+
+| Codec       | <i>a</i> | <i>b</i> | <i>c</i> | <i>d</i> |
+|-------------|----------|----------|----------|----------|
+| H.264       | 4.4344   | -1.7058  | 4.9654   | -4.1203  |
+| H.264-10bit | 4.6467   | -0.8091  | 5.9835   | -4.4398  |
+| H.265       | 4.3789   | -1.0208  | 5.7572   | -4.5625  |
+| H.265-10bit | 4.5458   | -0.866   | 6.1116   | -3.3828  |
+| VP9         | 4.3404   | -0.9961  | 4.5282   | -3.9641  |
+
+**Table 6 – Mode 3 – MO/TA**
+
+| Codec       | <i>a</i> | <i>b</i> | <i>c</i> | <i>d</i> |
+|-------------|----------|----------|----------|----------|
+| H.264       | 4.4365   | −1.4909  | 5.4251   | −4.5198  |
+| H.264-10bit | 4.5399   | −0.414   | 6.2249   | −4.2599  |
+| H.265       | 4.3089   | −0.6685  | 6.0551   | −4.6974  |
+| H.265-10bit | 4.9999   | −2.6821  | 1.5069   | −1.7664  |
+| VP9         | 4.4024   | −1.2504  | 2.9268   | −3.0087  |
+
+**Table 7 – Resolution upscaling**
+
+| End-device | <i>x</i> | <i>y</i> |
+|------------|----------|----------|
+| PC/TV      | −9.5497  | 1.1999   |
+| MO/TA      | −8.4690  | 1.1999   |
+
+**Table 8 – Frame rate upscaling**
+
+| End-device | <i>k</i> | <i>z</i> |
+|------------|----------|----------|
+| PC/TV      | 4.1696   | −8.3084  |
+| MO/TA      | 4.2701   | −6.3648  |
+
+## 8.2 Machine-learning-based part of the model
+
+The proposed random forest model estimates a residual prediction, i.e., the difference between the real video quality score obtained from subjective tests during model training and the prediction of the parametric part of the model, which uses only *QP* and the separate components addressing upscaling and temporal degradation due to the given frame rate. This difference can be explained by the contribution of features to the overall quality score, which are not available in the parametric model part.
+
+Different statistical aggregations of the features are computed and used as the input to the random forest model. In addition to the content-related features, the random forest model explicitly takes into account the prediction from the parametric part of the model as further input. The final random forest-based prediction is the summation of the prediction of the parametric part and the predicted residual.
+
+$$M_{\text{randomForest}} = M_{\text{parametric}} + \text{Residual} \quad (13)$$
+
+### 8.2.1 Random forest features
+
+This clause lists the features used in the random forest model. To generate features for the random forest model that are not directly available from the model input, aggregations of input data may be performed. The features are listed in Table 9.
+
+**Table 9 – Features**
+
+| Aggregated feature                                                                            | Type  | Feature index in code |
+|-----------------------------------------------------------------------------------------------|-------|-----------------------|
+| Minimum standard deviation of motion in the <i>x</i> -direction (horizontal motion) per frame | float | <i>x</i> [0]          |
+| Maximum frame size in bytes                                                                   | int   | <i>x</i> [1]          |
+| Mean bitrate per segment in kilobits per second                                               | float | <i>x</i> [2]          |
+
+**Table 9 – Features**
+
+| Aggregated feature                                                                                                                                  | Type                      | Feature index in code |
+|-----------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------|-----------------------|
+| Frame rate                                                                                                                                          | float                     | x[3]                  |
+| Resolution ( <i>width * height</i> ) of the distorted video                                                                                         | int                       | x[4]                  |
+| H.264 See Note                                                                                                                                      | boolean (0=False, 1=True) | x[5]                  |
+| H.264_10bit See Note                                                                                                                                | boolean (0=False, 1=True) | x[6]                  |
+| H.265 See Note                                                                                                                                      | boolean (0=False, 1=True) | x[7]                  |
+| H.265_10bit See Note                                                                                                                                | boolean (0=False, 1=True) | x[8]                  |
+| Interquartile range (IQR) of the average quantization parameter of non-I-frames                                                                     | float                     | x[9]                  |
+| IQR of the minimum quantization parameter per frame                                                                                                 | float                     | x[10]                 |
+| Kurtosis of the average motion per frame over all frames in a segment                                                                               | float                     | x[11]                 |
+| Kurtosis of the average quantization parameter of non-I-frames                                                                                      | float                     | x[12]                 |
+| Kurtosis of the non-I frame sizes                                                                                                                   | float                     | x[13]                 |
+| Mean of the average quantization parameter of non-I-frames                                                                                          | float                     | x[14]                 |
+| $M_{\text{parametric}}$                                                                                                                             | float                     | x[15]                 |
+| $Quant \left( Quant = \frac{QP_{\text{non-I-frames}}}{QP_{\text{max}}} \right)$                                                                     | float                     | x[16]                 |
+| Standard deviation of frame size of non-I frame in bits                                                                                             | float                     | x[17]                 |
+| Standard deviation of maximum QP of non-I frames                                                                                                    | float                     | x[18]                 |
+| VP9 See Note                                                                                                                                        | boolean (0=False, 1=True) | x[19]                 |
+| NOTE – A binary feature, e.g., in the case of [ITU-T H.264], this value is true (1) if the video is encoded with[ ITU-T H.264], otherwise false (0) |                           |                       |
+
+The random forest model uses 20 trees with a fixed depth of eight. The individual trees are transformed as functions and a function *predict* for aggregation (mean value of all individual tree predictions) of these trees for final prediction are added in the software attachment to this Recommendation as Python code, providing two different versions for MO/TA and PC/TV.
+
+The feature index in code column in Table 9 refers to how the aggregated features are indexed in the function *predict* with respect to the feature vector x.
+
+NOTE – Depending on the final random forest model used, feature indices used by the trees can be different and are selected in the corresponding *predict* function.
+
+## 8.3 Final prediction
+
+The final prediction of quality is the weighted average of the prediction from the parametric part and the random forest part.
+
+$$Q = w_1 * M_{\text{parametric}} + w_2 * M_{\text{randomForest}} \quad (14)$$
+
+Here,  $w_1 = 0.5$  and  $w_2 = 0.5$ ; both parts get equal importance in the final score.
+
+A final adjustment to the prediction in Equation 14 is added to compensate for differences in subjective ratings due to the heterogeneity of tests across different laboratories for the training and validation databases. The final per-media chunk prediction  $O_{27}$  is then given by
+
+$$O.27 = a * Q + b \quad (15)$$
+
+Here,  $a = 1.036$  and  $b = -0.1457$
+
+## 8.4 Per-second score prediction
+
+In addition to the overall video quality score, the model also outputs the per-second scores. The per-second video quality score ( $O.22$ ) is calculated as follows:
+
+$$O.22 = \frac{\text{mean}(QP_{\text{non-I,per-seg}})}{\text{mean}(QP_{\text{non-I,per-sec}})} * Q \quad (16)$$
+
+where
+
+$QP_{\text{non-I,per-seg}}$  is the average QP of all non-I frames in a segment;
+
+$QP_{\text{non-I,per-sec}}$  is the average QP of all non-I frames for each second;
+
+$Q$  is the per-segment video quality score as described in Equation (14).
+
+# Annex A
+
+## Helper function definitions
+
+(This annex forms an integral part of this Recommendation.)
+
+*MOSfromR* can be expressed as follows:
+
+```
+function MOSfromR(Q):
+    MOS_MAX = 4.5
+    MOS_MIN = 1.0
+
+    if Q >= 100:
+        return MOS_MAX
+    if Q <= 0:
+        return MOS_MIN
+
+    return (
+        MOS_MIN
+        + ((MOS_MAX - MOS_MIN) * Q / 100)
+        + Q * (Q - 60) * (100 - Q) * 0.000007
+    )
+```
+
+*RfromMOS* can be expressed as follows:
+
+```
+function RfromMOS(MOS):
+    x = (18566 - 6750 * MOS)
+    if MOS > 4.5:
+        MOS = 4.5
+
+    if x < 0:
+        num = 15 * sqrt(-903522 + 1113960 * MOS - 202500 * MOS * MOS)
+        den = 6750 * MOS - 18566
+        fra = num / den
+        h = (pi - atan(fra)) / 3
+    else:
+        num = 15 * sqrt(-903522 + 1113960 * MOS - 202500 * MOS * MOS)
+        den = 18566 - 6750 * MOS
+        fra = num / den
+        ar = atan(fra)
+        h = atan(num / den) / 3
+    R = 20.0 * (8 - sqrt(226) * cos(h + pi / 3)) / 3
+    return
+```
+
+*scaleto5* can be expressed as follows:
+
+```
+function scaleto5(x):
+
+    input_start = 1
+    input_end = 4.5
+    output_start = 1
+    output_end = 5
+
+    if x >= 4.5:
+        return 5
+```
+
+```
+    return output_start + ((output_end - output_start) /  
+(input_end - input_start)) * (  
+        x - input_start  
+)
+```
+
+# Appendix I
+
+## Performance figures
+
+(This appendix does not form an integral part of this Recommendation.)
+
+In this appendix, the root mean square errors (RMSEs) of Pv models are reported. Note that the numbers are reported after a final per-database mapping between the model output and the subjective scores of a database. This linear mapping is used to account for scale and bias variations between different databases.
+
+**Table I.1 – Validation performance of Pv model: The submitted model is the model trained on the exchanged training databases and frozen before creation of validation data. Models were retrained using a five-fold cross-validation approach, with their validation performance listed to show stability of the performance indicating no over-fitting.**
+
+|                         |                            |              |       |       |       |       |
+|-------------------------|----------------------------|--------------|-------|-------|-------|-------|
+| <b>Bitstream mode 3</b> | Submitted model            | 0.421        |       |       |       |       |
+|                         | Five-fold cross-validation | <b>0.394</b> | 0.407 | 0.402 | 0.413 | 0.401 |
+
+The re-training of the submitted model was performed on five different splits. The splits were defined on the database level. The following is the procedure that was followed to determine the splits.
+
+- All training and validation databases were merged to obtain in total 26 different short databases (18 PC/TV and eight MO/TA).
+- A level of difficulty of prediction for each database was determined based on average prediction error over all models.
+- A 50:50 training:validation split was determined randomly, but respecting the level of difficulty. In total, five different splits were defined. Each split had a balanced distribution of databases based on difficulty in both the training and validation.
+- The 50:50 split was separately performed for PC/TV and MO/TA cases.
+- The final model coefficients correspond to the best performing split.
+
+# Bibliography
+
+- [b-ITU-T G.1022] Recommendation ITU-T G.1022 (2016), *Buffer models for media streams on TCP transport*.
+- [b-ITU-T P.800.1] Recommendation ITU-T P.800.1 (2016), *Mean opinion score (MOS) terminology*.
+- [b-ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*.
+- [b-ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Lower resolution application area*.
+- [b-ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Higher resolution application area*.
+- [b-ITU-T P.1202] Recommendation ITU-T P.1202 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality*.
+- [b-ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – Lower resolution application area*.
+- [b-ITU-T P.1203] Recommendation ITU-T P.1203 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport*.
+- [b-ITU-T P.1203.2] Recommendation ITU-T P.1203.2 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module*.
+- [b-ITU-T P.1204.4] Recommendation ITU-T P.1204.4 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full and reduced reference pixel information*.
+- [b-ITU-T P.1204.5] Recommendation ITU-T P.1204.5 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to transport and received pixel information*.
+- [b-ITU-T P.1401] Recommendation ITU-T P.1401 (2020), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*.
+
+## SERIES OF ITU-T RECOMMENDATIONS
+
+|                 |                                                                                                                                                           |
+|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------|
+| Series A        | Organization of the work of ITU-T                                                                                                                         |
+| Series D        | Tariff and accounting principles and international telecommunication/ICT economic and policy issues                                                       |
+| Series E        | Overall network operation, telephone service, service operation and human factors                                                                         |
+| Series F        | Non-telephone telecommunication services                                                                                                                  |
+| Series G        | Transmission systems and media, digital systems and networks                                                                                              |
+| Series H        | Audiovisual and multimedia systems                                                                                                                        |
+| Series I        | Integrated services digital network                                                                                                                       |
+| Series J        | Cable networks and transmission of television, sound programme and other multimedia signals                                                               |
+| Series K        | Protection against interference                                                                                                                           |
+| Series L        | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant |
+| Series M        | Telecommunication management, including TMN and network maintenance                                                                                       |
+| Series N        | Maintenance: international sound programme and television transmission circuits                                                                           |
+| Series O        | Specifications of measuring equipment                                                                                                                     |
+| <b>Series P</b> | <b>Telephone transmission quality, telephone installations, local line networks</b>                                                                       |
+| Series Q        | Switching and signalling, and associated measurements and tests                                                                                           |
+| Series R        | Telegraph transmission                                                                                                                                    |
+| Series S        | Telegraph services terminal equipment                                                                                                                     |
+| Series T        | Terminals for telematic services                                                                                                                          |
+| Series U        | Telegraph switching                                                                                                                                       |
+| Series V        | Data communication over the telephone network                                                                                                             |
+| Series X        | Data networks, open system communications and security                                                                                                    |
+| Series Y        | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities                               |
+| Series Z        | Languages and general software aspects for telecommunication systems                                                                                      |