File size: 11,246 Bytes
d520909 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 |
"""
Image Cropping Utilities
Functions for extracting and managing region crops from document images.
"""
import hashlib
import logging
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple, Union
import numpy as np
from PIL import Image
from ..chunks.models import BoundingBox, DocumentChunk
logger = logging.getLogger(__name__)
def crop_region(
image: Union[np.ndarray, Image.Image],
bbox: BoundingBox,
padding_percent: float = 0.02,
) -> np.ndarray:
"""
Crop a region from an image.
Args:
image: Source image (numpy array or PIL Image)
bbox: Bounding box to crop (can be normalized or pixel)
padding_percent: Padding to add around the crop (0-1)
Returns:
Cropped image as numpy array
"""
# Convert to numpy if needed
if isinstance(image, Image.Image):
image = np.array(image)
height, width = image.shape[:2]
# Convert to pixel coordinates if normalized
if bbox.normalized:
pixel_bbox = bbox.to_pixel(width, height)
else:
pixel_bbox = bbox
# Apply padding
pad_x = int(pixel_bbox.width * padding_percent)
pad_y = int(pixel_bbox.height * padding_percent)
x_min = max(0, int(pixel_bbox.x_min) - pad_x)
y_min = max(0, int(pixel_bbox.y_min) - pad_y)
x_max = min(width, int(pixel_bbox.x_max) + pad_x)
y_max = min(height, int(pixel_bbox.y_max) + pad_y)
# Ensure valid crop region
if x_max <= x_min or y_max <= y_min:
logger.warning(f"Invalid crop region: ({x_min}, {y_min}, {x_max}, {y_max})")
return np.zeros((1, 1, 3), dtype=np.uint8)
return image[y_min:y_max, x_min:x_max].copy()
def crop_chunk(
image: Union[np.ndarray, Image.Image],
chunk: DocumentChunk,
padding_percent: float = 0.02,
) -> np.ndarray:
"""
Crop the region corresponding to a chunk.
Args:
image: Page image
chunk: Document chunk with bbox
padding_percent: Padding around crop
Returns:
Cropped image
"""
return crop_region(image, chunk.bbox, padding_percent)
def crop_multiple_regions(
image: Union[np.ndarray, Image.Image],
bboxes: List[BoundingBox],
padding_percent: float = 0.02,
) -> List[np.ndarray]:
"""
Crop multiple regions from an image.
Args:
image: Source image
bboxes: List of bounding boxes
padding_percent: Padding around crops
Returns:
List of cropped images
"""
return [crop_region(image, bbox, padding_percent) for bbox in bboxes]
class CropManager:
"""
Manages crop extraction and storage.
Provides caching and organized storage for document crops.
"""
def __init__(
self,
output_dir: Union[str, Path],
format: str = "png",
quality: int = 95,
):
self.output_dir = Path(output_dir)
self.format = format.lower()
self.quality = quality
self._cache: Dict[str, str] = {}
# Ensure output directory exists
self.output_dir.mkdir(parents=True, exist_ok=True)
def get_crop_path(
self,
doc_id: str,
page: int,
bbox: BoundingBox,
) -> Path:
"""Generate a path for a crop."""
# Create stable filename from bbox
bbox_str = f"{bbox.x_min:.4f}_{bbox.y_min:.4f}_{bbox.x_max:.4f}_{bbox.y_max:.4f}"
bbox_hash = hashlib.md5(bbox_str.encode()).hexdigest()[:8]
filename = f"{doc_id}_p{page}_{bbox_hash}.{self.format}"
return self.output_dir / doc_id / filename
def save_crop(
self,
image: Union[np.ndarray, Image.Image],
doc_id: str,
page: int,
bbox: BoundingBox,
padding_percent: float = 0.02,
) -> str:
"""
Crop and save a region.
Args:
image: Source page image
doc_id: Document ID
page: Page number
bbox: Region to crop
padding_percent: Padding around crop
Returns:
Path to saved crop
"""
# Check cache
cache_key = f"{doc_id}_{page}_{bbox.xyxy}"
if cache_key in self._cache:
return self._cache[cache_key]
# Crop region
crop = crop_region(image, bbox, padding_percent)
# Convert to PIL
pil_crop = Image.fromarray(crop)
# Ensure directory exists
crop_path = self.get_crop_path(doc_id, page, bbox)
crop_path.parent.mkdir(parents=True, exist_ok=True)
# Save
if self.format == "jpg" or self.format == "jpeg":
pil_crop.save(crop_path, format="JPEG", quality=self.quality)
else:
pil_crop.save(crop_path, format=self.format.upper())
# Cache
path_str = str(crop_path)
self._cache[cache_key] = path_str
return path_str
def save_chunk_crop(
self,
image: Union[np.ndarray, Image.Image],
chunk: DocumentChunk,
padding_percent: float = 0.02,
) -> str:
"""
Save crop for a document chunk.
Args:
image: Page image
chunk: Chunk to crop
padding_percent: Padding around crop
Returns:
Path to saved crop
"""
return self.save_crop(
image=image,
doc_id=chunk.doc_id,
page=chunk.page,
bbox=chunk.bbox,
padding_percent=padding_percent,
)
def get_cached_crop(
self,
doc_id: str,
page: int,
bbox: BoundingBox,
) -> Optional[str]:
"""Get path to cached crop if it exists."""
cache_key = f"{doc_id}_{page}_{bbox.xyxy}"
return self._cache.get(cache_key)
def load_crop(self, path: Union[str, Path]) -> Optional[np.ndarray]:
"""Load a crop from disk."""
path = Path(path)
if not path.exists():
return None
try:
img = Image.open(path)
return np.array(img)
except Exception as e:
logger.warning(f"Failed to load crop {path}: {e}")
return None
def clear_cache(self) -> None:
"""Clear the path cache."""
self._cache.clear()
def cleanup_doc(self, doc_id: str) -> int:
"""
Remove all crops for a document.
Returns number of files removed.
"""
doc_dir = self.output_dir / doc_id
if not doc_dir.exists():
return 0
count = 0
for crop_file in doc_dir.glob(f"*.{self.format}"):
try:
crop_file.unlink()
count += 1
except Exception:
pass
# Remove directory if empty
try:
doc_dir.rmdir()
except OSError:
pass
# Clear cache entries
self._cache = {
k: v for k, v in self._cache.items()
if not k.startswith(f"{doc_id}_")
}
return count
def create_annotated_image(
image: Union[np.ndarray, Image.Image],
bboxes: List[BoundingBox],
labels: Optional[List[str]] = None,
colors: Optional[List[Tuple[int, int, int]]] = None,
line_width: int = 2,
font_size: int = 12,
) -> np.ndarray:
"""
Create an annotated image with bounding boxes.
Args:
image: Source image
bboxes: Bounding boxes to draw
labels: Optional labels for each box
colors: Optional colors for each box (RGB tuples)
line_width: Line width for boxes
font_size: Font size for labels
Returns:
Annotated image as numpy array
"""
from PIL import ImageDraw, ImageFont
# Convert to PIL
if isinstance(image, np.ndarray):
pil_image = Image.fromarray(image).copy()
else:
pil_image = image.copy()
draw = ImageDraw.Draw(pil_image)
width, height = pil_image.size
# Default colors - rotating palette
default_colors = [
(255, 0, 0), # Red
(0, 255, 0), # Green
(0, 0, 255), # Blue
(255, 255, 0), # Yellow
(255, 0, 255), # Magenta
(0, 255, 255), # Cyan
(255, 128, 0), # Orange
(128, 0, 255), # Purple
]
# Try to load font
try:
font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf", font_size)
except Exception:
font = ImageFont.load_default()
for i, bbox in enumerate(bboxes):
# Get color
if colors and i < len(colors):
color = colors[i]
else:
color = default_colors[i % len(default_colors)]
# Convert to pixels if normalized
if bbox.normalized:
x_min = int(bbox.x_min * width)
y_min = int(bbox.y_min * height)
x_max = int(bbox.x_max * width)
y_max = int(bbox.y_max * height)
else:
x_min = int(bbox.x_min)
y_min = int(bbox.y_min)
x_max = int(bbox.x_max)
y_max = int(bbox.y_max)
# Draw rectangle
draw.rectangle(
[(x_min, y_min), (x_max, y_max)],
outline=color,
width=line_width,
)
# Draw label if provided
if labels and i < len(labels):
label = labels[i]
# Draw label background
text_bbox = draw.textbbox((x_min, y_min - font_size - 4), label, font=font)
draw.rectangle(text_bbox, fill=color)
# Draw text
draw.text(
(x_min, y_min - font_size - 4),
label,
fill=(255, 255, 255),
font=font,
)
return np.array(pil_image)
def highlight_region(
image: Union[np.ndarray, Image.Image],
bbox: BoundingBox,
highlight_color: Tuple[int, int, int] = (255, 255, 0),
opacity: float = 0.3,
) -> np.ndarray:
"""
Highlight a region in an image with semi-transparent overlay.
Args:
image: Source image
bbox: Region to highlight
highlight_color: Color for highlight (RGB)
opacity: Opacity of highlight (0-1)
Returns:
Image with highlighted region
"""
# Convert to numpy
if isinstance(image, Image.Image):
img_array = np.array(image).copy()
else:
img_array = image.copy()
height, width = img_array.shape[:2]
# Convert to pixels if normalized
if bbox.normalized:
x_min = int(bbox.x_min * width)
y_min = int(bbox.y_min * height)
x_max = int(bbox.x_max * width)
y_max = int(bbox.y_max * height)
else:
x_min = int(bbox.x_min)
y_min = int(bbox.y_min)
x_max = int(bbox.x_max)
y_max = int(bbox.y_max)
# Clip to valid range
x_min = max(0, x_min)
y_min = max(0, y_min)
x_max = min(width, x_max)
y_max = min(height, y_max)
# Create overlay
overlay = np.full((y_max - y_min, x_max - x_min, 3), highlight_color, dtype=np.uint8)
# Blend with original
region = img_array[y_min:y_max, x_min:x_max]
blended = (region * (1 - opacity) + overlay * opacity).astype(np.uint8)
img_array[y_min:y_max, x_min:x_max] = blended
return img_array
|