SceneWeaver / inpainting_blender.py
DawnC's picture
Upload 15 files
991a517 verified
raw
history blame
16.5 kB
import logging
from typing import Any, Dict, Optional, Tuple
import cv2
import numpy as np
from PIL import Image
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)
class InpaintingBlender:
"""
Handles mask processing, prompt enhancement, and result blending for inpainting.
This class encapsulates all pre-processing and post-processing operations
needed for inpainting, separate from the main generation pipeline.
Attributes:
min_mask_coverage: Minimum mask coverage threshold
max_mask_coverage: Maximum mask coverage threshold
Example:
>>> blender = InpaintingBlender()
>>> processed_mask, info = blender.prepare_mask(mask, (512, 512), feather_radius=8)
>>> enhanced_prompt, negative = blender.enhance_prompt("a flower", image, mask)
>>> result = blender.blend_result(original, generated, mask)
"""
def __init__(
self,
min_mask_coverage: float = 0.01,
max_mask_coverage: float = 0.95
):
"""
Initialize the InpaintingBlender.
Parameters
----------
min_mask_coverage : float
Minimum mask coverage (default: 1%)
max_mask_coverage : float
Maximum mask coverage (default: 95%)
"""
self.min_mask_coverage = min_mask_coverage
self.max_mask_coverage = max_mask_coverage
logger.info("InpaintingBlender initialized")
def prepare_mask(
self,
mask: Image.Image,
target_size: Tuple[int, int],
feather_radius: int = 8
) -> Tuple[Image.Image, Dict[str, Any]]:
"""
Prepare and validate mask for inpainting.
Parameters
----------
mask : PIL.Image
Input mask (white = inpaint area)
target_size : tuple
Target (width, height) to match input image
feather_radius : int
Feathering radius in pixels
Returns
-------
tuple
(processed_mask, validation_info)
Raises
------
ValueError
If mask coverage is outside acceptable range
"""
# Convert to grayscale
if mask.mode != 'L':
mask = mask.convert('L')
# Resize to match target
if mask.size != target_size:
mask = mask.resize(target_size, Image.LANCZOS)
# Convert to array for processing
mask_array = np.array(mask)
# Calculate coverage
total_pixels = mask_array.size
white_pixels = np.count_nonzero(mask_array > 127)
coverage = white_pixels / total_pixels
validation_info = {
"coverage": coverage,
"white_pixels": white_pixels,
"total_pixels": total_pixels,
"feather_radius": feather_radius,
"valid": True,
"warning": ""
}
# Validate coverage
if coverage < self.min_mask_coverage:
validation_info["valid"] = False
validation_info["warning"] = (
f"Mask coverage too low ({coverage:.1%}). "
f"Please select a larger area to inpaint."
)
logger.warning(f"Mask coverage {coverage:.1%} below minimum {self.min_mask_coverage:.1%}")
elif coverage > self.max_mask_coverage:
validation_info["valid"] = False
validation_info["warning"] = (
f"Mask coverage too high ({coverage:.1%}). "
f"Consider using background generation instead."
)
logger.warning(f"Mask coverage {coverage:.1%} above maximum {self.max_mask_coverage:.1%}")
# Apply feathering
if feather_radius > 0:
mask_array = cv2.GaussianBlur(
mask_array,
(feather_radius * 2 + 1, feather_radius * 2 + 1),
feather_radius / 2
)
logger.debug(f"Applied {feather_radius}px feathering to mask")
processed_mask = Image.fromarray(mask_array, mode='L')
return processed_mask, validation_info
def enhance_prompt_for_inpainting(
self,
prompt: str,
image: Image.Image,
mask: Image.Image
) -> Tuple[str, str]:
"""
Enhance prompt based on non-masked region analysis.
Analyzes the surrounding context to generate appropriate
lighting and color descriptors.
Parameters
----------
prompt : str
User-provided prompt
image : PIL.Image
Original image
mask : PIL.Image
Inpainting mask
Returns
-------
tuple
(enhanced_prompt, negative_prompt)
"""
logger.info("Enhancing prompt for inpainting context...")
# Convert to arrays
img_array = np.array(image.convert('RGB'))
mask_array = np.array(mask.convert('L'))
# Analyze non-masked regions
non_masked = mask_array < 127
if not np.any(non_masked):
# No context available
enhanced_prompt = f"{prompt}, high quality, detailed, photorealistic"
negative_prompt = self._get_inpainting_negative_prompt()
return enhanced_prompt, negative_prompt
# Extract context pixels
context_pixels = img_array[non_masked]
# Convert to Lab for analysis
context_lab = cv2.cvtColor(
context_pixels.reshape(-1, 1, 3),
cv2.COLOR_RGB2LAB
).reshape(-1, 3)
# Use robust statistics (median) to avoid outlier influence
median_l = np.median(context_lab[:, 0])
median_b = np.median(context_lab[:, 2])
# Analyze lighting conditions
lighting_descriptors = []
if median_l > 170:
lighting_descriptors.append("bright")
elif median_l > 130:
lighting_descriptors.append("well-lit")
elif median_l > 80:
lighting_descriptors.append("moderate lighting")
else:
lighting_descriptors.append("dim lighting")
# Analyze color temperature (b channel: blue(-) to yellow(+))
if median_b > 140:
lighting_descriptors.append("warm golden tones")
elif median_b > 120:
lighting_descriptors.append("warm afternoon light")
elif median_b < 110:
lighting_descriptors.append("cool neutral tones")
# Calculate saturation from context
hsv = cv2.cvtColor(context_pixels.reshape(-1, 1, 3), cv2.COLOR_RGB2HSV)
median_saturation = np.median(hsv[:, :, 1])
if median_saturation > 150:
lighting_descriptors.append("vibrant colors")
elif median_saturation < 80:
lighting_descriptors.append("subtle muted colors")
# Build enhanced prompt
lighting_desc = ", ".join(lighting_descriptors) if lighting_descriptors else ""
quality_suffix = "high quality, detailed, photorealistic, seamless integration"
if lighting_desc:
enhanced_prompt = f"{prompt}, {lighting_desc}, {quality_suffix}"
else:
enhanced_prompt = f"{prompt}, {quality_suffix}"
negative_prompt = self._get_inpainting_negative_prompt()
logger.info(f"Enhanced prompt with context: {lighting_desc}")
return enhanced_prompt, negative_prompt
def _get_inpainting_negative_prompt(self) -> str:
"""Get standard negative prompt for inpainting."""
return (
"inconsistent lighting, wrong perspective, mismatched colors, "
"visible seams, blending artifacts, color bleeding, "
"blurry, low quality, distorted, deformed, "
"harsh edges, unnatural transition"
)
def blend_result(
self,
original: Image.Image,
generated: Image.Image,
mask: Image.Image
) -> Image.Image:
"""
Blend generated content with original image.
Uses color matching and linear color space blending for seamless results.
Parameters
----------
original : PIL.Image
Original image
generated : PIL.Image
Generated inpainted image
mask : PIL.Image
Blending mask (white = use generated)
Returns
-------
PIL.Image
Blended result
"""
logger.info("Blending inpainting result with color matching...")
# Ensure same size
if generated.size != original.size:
generated = generated.resize(original.size, Image.LANCZOS)
if mask.size != original.size:
mask = mask.resize(original.size, Image.LANCZOS)
# Convert to arrays
orig_array = np.array(original.convert('RGB')).astype(np.float32)
gen_array = np.array(generated.convert('RGB')).astype(np.float32)
mask_array = np.array(mask.convert('L')).astype(np.float32) / 255.0
# Apply color matching to generated region (use original mask for accurate boundary detection)
gen_array = self._match_colors_at_boundary(orig_array, gen_array, mask_array)
# Create blend mask: soften edges ONLY for blending (not for generation)
# This ensures full generation coverage while smooth blending at edges
blend_mask = self._create_blend_mask(mask_array)
# sRGB to linear conversion
def srgb_to_linear(img: np.ndarray) -> np.ndarray:
img_norm = img / 255.0
return np.where(
img_norm <= 0.04045,
img_norm / 12.92,
np.power((img_norm + 0.055) / 1.055, 2.4)
)
def linear_to_srgb(img: np.ndarray) -> np.ndarray:
img_clipped = np.clip(img, 0, 1)
return np.where(
img_clipped <= 0.0031308,
12.92 * img_clipped,
1.055 * np.power(img_clipped, 1/2.4) - 0.055
)
# Convert to linear space
orig_linear = srgb_to_linear(orig_array)
gen_linear = srgb_to_linear(gen_array)
# Alpha blending in linear space using the blend mask (with softened edges)
alpha = blend_mask[:, :, np.newaxis]
result_linear = gen_linear * alpha + orig_linear * (1 - alpha)
# Convert back to sRGB
result_srgb = linear_to_srgb(result_linear)
result_array = (result_srgb * 255).astype(np.uint8)
logger.debug("Blending completed with color matching")
return Image.fromarray(result_array)
def _match_colors_at_boundary(
self,
original: np.ndarray,
generated: np.ndarray,
mask: np.ndarray
) -> np.ndarray:
"""
Match colors of generated content to original at the boundary.
Uses histogram matching in Lab color space for natural blending.
Parameters
----------
original : np.ndarray
Original image array (float32, 0-255)
generated : np.ndarray
Generated image array (float32, 0-255)
mask : np.ndarray
Mask array (float32, 0-1)
Returns
-------
np.ndarray
Color-matched generated image
"""
# Create boundary region mask (dilated mask - eroded mask)
mask_binary = (mask > 0.5).astype(np.uint8) * 255
# Create narrow boundary region for sampling original colors
kernel_size = 25 # Pixels to sample around boundary
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (kernel_size, kernel_size))
dilated = cv2.dilate(mask_binary, kernel, iterations=1)
eroded = cv2.erode(mask_binary, kernel, iterations=1)
# Outer boundary (original side)
outer_boundary = (dilated > 0) & (mask_binary == 0)
# Inner boundary (generated side)
inner_boundary = (mask_binary > 0) & (eroded == 0)
if not np.any(outer_boundary) or not np.any(inner_boundary):
logger.debug("No boundary region found, skipping color matching")
return generated
# Convert to Lab color space
orig_lab = cv2.cvtColor(original.astype(np.uint8), cv2.COLOR_RGB2LAB).astype(np.float32)
gen_lab = cv2.cvtColor(generated.astype(np.uint8), cv2.COLOR_RGB2LAB).astype(np.float32)
# Sample colors from boundary regions
orig_boundary_pixels = orig_lab[outer_boundary]
gen_boundary_pixels = gen_lab[inner_boundary]
if len(orig_boundary_pixels) < 10 or len(gen_boundary_pixels) < 10:
logger.debug("Not enough boundary pixels, skipping color matching")
return generated
# Calculate statistics
orig_mean = np.mean(orig_boundary_pixels, axis=0)
orig_std = np.std(orig_boundary_pixels, axis=0) + 1e-6
gen_mean = np.mean(gen_boundary_pixels, axis=0)
gen_std = np.std(gen_boundary_pixels, axis=0) + 1e-6
# Calculate correction factors
# Only correct L (lightness) and a,b (color) channels
l_correction = (orig_mean[0] - gen_mean[0]) * 0.7 # 70% correction for lightness
a_correction = (orig_mean[1] - gen_mean[1]) * 0.5 # 50% correction for color
b_correction = (orig_mean[2] - gen_mean[2]) * 0.5
logger.debug(f"Color correction: L={l_correction:.1f}, a={a_correction:.1f}, b={b_correction:.1f}")
# Apply correction to masked region only
corrected_lab = gen_lab.copy()
mask_region = mask > 0.3 # Apply to most of masked region
corrected_lab[mask_region, 0] = np.clip(
corrected_lab[mask_region, 0] + l_correction, 0, 255
)
corrected_lab[mask_region, 1] = np.clip(
corrected_lab[mask_region, 1] + a_correction, 0, 255
)
corrected_lab[mask_region, 2] = np.clip(
corrected_lab[mask_region, 2] + b_correction, 0, 255
)
# Convert back to RGB
corrected_rgb = cv2.cvtColor(
corrected_lab.astype(np.uint8),
cv2.COLOR_LAB2RGB
).astype(np.float32)
logger.info("Applied boundary color matching")
return corrected_rgb
def _create_blend_mask(self, mask: np.ndarray) -> np.ndarray:
"""
Create a blend mask with softened edges for natural compositing.
The mask interior stays fully opaque (1.0) while only the edges
get a smooth transition. This preserves full generated content
while blending naturally at boundaries.
Parameters
----------
mask : np.ndarray
Original mask array (float32, 0-1)
Returns
-------
np.ndarray
Blend mask with soft edges but solid interior
"""
# Convert to uint8 for morphological operations
mask_uint8 = (mask * 255).astype(np.uint8)
# Create eroded version (solid interior)
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (15, 15))
eroded = cv2.erode(mask_uint8, kernel, iterations=1)
# Create smooth transition zone at edges only
# Blur the original mask for edge softness
blurred = cv2.GaussianBlur(mask_uint8, (15, 15), 4)
# Combine: use eroded (solid) for interior, blurred for edges
# Where eroded > 0, use full opacity; elsewhere use blurred transition
result = np.where(eroded > 128, mask_uint8, blurred)
# Final light smoothing
result = cv2.GaussianBlur(result, (5, 5), 1)
# Convert back to float
blend_mask = result.astype(np.float32) / 255.0
logger.debug("Created blend mask with soft edges and solid interior")
return blend_mask
def validate_inputs(
self,
image: Image.Image,
mask: Image.Image
) -> Tuple[bool, str]:
"""
Validate image and mask inputs before processing.
Parameters
----------
image : PIL.Image
Input image
mask : PIL.Image
Input mask
Returns
-------
tuple
(is_valid, error_message)
"""
if image is None:
return False, "No image provided"
if mask is None:
return False, "No mask provided"
# Check sizes match
if image.size != mask.size:
# Will be resized later, so just log a warning
logger.warning(f"Image size {image.size} != mask size {mask.size}, will resize")
return True, ""