Spaces:
Runtime error
Runtime error
File size: 9,503 Bytes
0ba6002 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 | """
Image utility functions for preprocessing
Provides functions for resizing, normalization, and quality validation
of card images.
"""
import cv2
import numpy as np
from typing import Dict, Tuple
from ..utils.logger import get_logger
logger = get_logger(__name__)
def resize_image(image: np.ndarray, size: int = 256) -> np.ndarray:
"""
Resize image to standard square size
Resizes the image to size×size pixels while maintaining quality.
Uses INTER_AREA for shrinking (better quality) and INTER_CUBIC
for enlarging.
Args:
image: Input image (H×W×C)
size: Target size in pixels (default: 256)
Returns:
Resized image (size×size×C)
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided to resize_image")
current_height, current_width = image.shape[:2]
# Choose interpolation method based on whether we're upscaling or downscaling
if current_height > size or current_width > size:
# Downscaling - use INTER_AREA for better quality
interpolation = cv2.INTER_AREA
else:
# Upscaling - use INTER_CUBIC for smoother results
interpolation = cv2.INTER_CUBIC
# Resize to square
resized = cv2.resize(image, (size, size), interpolation=interpolation)
logger.debug(f"Resized image from {current_width}×{current_height} to {size}×{size}")
return resized
def normalize_pixels(image: np.ndarray) -> np.ndarray:
"""
Normalize pixel values to [0, 1] range
Converts uint8 image (0-255) to float32 (0.0-1.0) for model processing.
Args:
image: Input image (uint8)
Returns:
Normalized image (float32 in range [0, 1])
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided to normalize_pixels")
# Convert to float32 and normalize to [0, 1]
normalized = image.astype(np.float32) / 255.0
logger.debug(f"Normalized image: min={normalized.min():.3f}, max={normalized.max():.3f}")
return normalized
def denormalize_pixels(image: np.ndarray) -> np.ndarray:
"""
Denormalize pixels from [0, 1] back to [0, 255]
Converts float32 image back to uint8 for display/saving.
Args:
image: Normalized image (float32 in range [0, 1])
Returns:
Denormalized image (uint8 in range [0, 255])
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided to denormalize_pixels")
# Convert to [0, 255] range and uint8
denormalized = (image * 255.0).clip(0, 255).astype(np.uint8)
return denormalized
def check_image_quality(
image: np.ndarray,
blur_threshold: float = 100.0,
brightness_range: Tuple[float, float] = (30.0, 225.0),
contrast_threshold: float = 30.0
) -> Dict[str, float]:
"""
Check image quality metrics (blur, brightness, contrast)
Analyzes the image to detect quality issues that could affect
feature extraction or classification.
Args:
image: Input image (uint8)
blur_threshold: Minimum blur score (Laplacian variance) for sharp image
brightness_range: Acceptable brightness range (min, max)
contrast_threshold: Minimum standard deviation for adequate contrast
Returns:
Dictionary with quality metrics:
- blur_score: Laplacian variance (higher = sharper)
- brightness: Mean pixel value (0-255)
- contrast: Standard deviation of pixels
- is_acceptable: Boolean indicating if image passes all checks
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided to check_image_quality")
# Convert to grayscale for analysis
if len(image.shape) == 3:
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
else:
gray = image
# 1. Blur detection using Laplacian variance
# Higher variance = sharper edges = less blur
laplacian = cv2.Laplacian(gray, cv2.CV_64F)
blur_score = laplacian.var()
# 2. Brightness (mean pixel value)
brightness = gray.mean()
# 3. Contrast (standard deviation of pixel values)
contrast = gray.std()
# Determine if image is acceptable
is_acceptable = (
blur_score >= blur_threshold and
brightness_range[0] <= brightness <= brightness_range[1] and
contrast >= contrast_threshold
)
quality_metrics = {
'blur_score': float(blur_score),
'brightness': float(brightness),
'contrast': float(contrast),
'is_acceptable': is_acceptable
}
if not is_acceptable:
logger.warning(
f"Image quality issues detected - "
f"blur: {blur_score:.1f} (threshold: {blur_threshold}), "
f"brightness: {brightness:.1f} (range: {brightness_range}), "
f"contrast: {contrast:.1f} (threshold: {contrast_threshold})"
)
return quality_metrics
def adaptive_histogram_equalization(image: np.ndarray, clip_limit: float = 2.0) -> np.ndarray:
"""
Apply CLAHE (Contrast Limited Adaptive Histogram Equalization)
Improves contrast in images with poor lighting conditions.
Args:
image: Input image (BGR)
clip_limit: Threshold for contrast limiting (default: 2.0)
Returns:
Contrast-enhanced image
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided")
# Convert to LAB color space
lab = cv2.cvtColor(image, cv2.COLOR_BGR2LAB)
# Split channels
l, a, b = cv2.split(lab)
# Apply CLAHE to L channel
clahe = cv2.createCLAHE(clipLimit=clip_limit, tileGridSize=(8, 8))
l_enhanced = clahe.apply(l)
# Merge channels
lab_enhanced = cv2.merge([l_enhanced, a, b])
# Convert back to BGR
enhanced = cv2.cvtColor(lab_enhanced, cv2.COLOR_LAB2BGR)
logger.debug("Applied adaptive histogram equalization")
return enhanced
def remove_noise(image: np.ndarray, kernel_size: int = 5) -> np.ndarray:
"""
Remove noise from image using bilateral filter
Smooths image while preserving edges.
Args:
image: Input image
kernel_size: Filter kernel size (default: 5)
Returns:
Denoised image
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided")
# Apply bilateral filter (preserves edges while smoothing)
denoised = cv2.bilateralFilter(image, kernel_size, 75, 75)
logger.debug(f"Applied bilateral filter with kernel size {kernel_size}")
return denoised
def auto_rotate_card(image: np.ndarray) -> Tuple[np.ndarray, float]:
"""
Automatically detect and correct card rotation
Detects if card is rotated and corrects to upright position.
Args:
image: Input card image
Returns:
Tuple of (rotated_image, rotation_angle_degrees)
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided")
# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Detect edges
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
# Detect lines using Hough transform
lines = cv2.HoughLines(edges, 1, np.pi / 180, 200)
if lines is None or len(lines) == 0:
logger.debug("No lines detected for rotation correction")
return image, 0.0
# Find dominant angle
angles = []
for line in lines:
rho, theta = line[0]
angle = np.degrees(theta)
angles.append(angle)
# Get median angle
median_angle = np.median(angles)
# Correct angle to [-45, 45] range
if median_angle > 135:
rotation_angle = median_angle - 180
elif median_angle > 45:
rotation_angle = median_angle - 90
else:
rotation_angle = median_angle
# Only rotate if angle is significant (> 2 degrees)
if abs(rotation_angle) < 2:
return image, 0.0
# Rotate image
height, width = image.shape[:2]
center = (width // 2, height // 2)
rotation_matrix = cv2.getRotationMatrix2D(center, rotation_angle, 1.0)
rotated = cv2.warpAffine(
image,
rotation_matrix,
(width, height),
flags=cv2.INTER_CUBIC,
borderMode=cv2.BORDER_REPLICATE
)
logger.debug(f"Rotated image by {rotation_angle:.2f} degrees")
return rotated, rotation_angle
def crop_to_content(image: np.ndarray, padding: int = 10) -> np.ndarray:
"""
Crop image to content (remove large uniform borders)
Args:
image: Input image
padding: Pixels to add around detected content (default: 10)
Returns:
Cropped image
"""
if image is None or image.size == 0:
raise ValueError("Empty or None image provided")
# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Threshold to find content
_, thresh = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY)
# Find bounding box of content
coords = cv2.findNonZero(thresh)
if coords is None:
logger.warning("No content found in image")
return image
x, y, w, h = cv2.boundingRect(coords)
# Add padding
height, width = image.shape[:2]
x = max(0, x - padding)
y = max(0, y - padding)
w = min(width - x, w + 2 * padding)
h = min(height - y, h + 2 * padding)
# Crop image
cropped = image[y:y+h, x:x+w]
logger.debug(f"Cropped to content: {w}×{h}")
return cropped
|