feat: extend covering detection to include black/redaction rectangles

Expands whiteout detection to handle:
- White rectangles (RGB >= 0.95) - correction tape / white-out
- Black rectangles (RGB <= 0.05) - redaction / censoring
- Other solid fills (very dark or very light) - potential covering

Adds color_type to covered text results for better logging.
Logs now show breakdown by cover type (white, black, other).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
egg
2025-12-04 07:34:35 +08:00
parent 63b474f93a
commit bc66f72352

View File

@@ -1894,14 +1894,20 @@ class DirectExtractionEngine:
if self.enable_whiteout_detection: if self.enable_whiteout_detection:
covered = self._detect_whiteout_covered_text(page, page_num) covered = self._detect_whiteout_covered_text(page, page_num)
result['covered_word_bboxes'] = [fitz.Rect(w['bbox']) for w in covered] result['covered_word_bboxes'] = [fitz.Rect(w['bbox']) for w in covered]
result['covered_words_detail'] = covered # Include color_type info
if covered: if covered:
logger.info(f"Page {page_num}: Detected {len(covered)} text regions covered by white-out") # Count by color type
white_covered = sum(1 for c in covered if c.get('color_type') == 'white')
black_covered = sum(1 for c in covered if c.get('color_type') == 'black')
other_covered = len(covered) - white_covered - black_covered
logger.info(f"Page {page_num}: Detected {len(covered)} covered text regions "
f"(white: {white_covered}, black/redaction: {black_covered}, other: {other_covered})")
return result return result
def _detect_whiteout_covered_text(self, page: fitz.Page, page_num: int) -> List[Dict]: def _detect_whiteout_covered_text(self, page: fitz.Page, page_num: int) -> List[Dict]:
""" """
Detect text covered by white rectangles ("white-out" / "correction tape" effect). Detect text covered by solid color rectangles (white-out, black redaction, or any solid fill).
Uses IoU (Intersection over Union) to determine if text is covered. Uses IoU (Intersection over Union) to determine if text is covered.
@@ -1910,30 +1916,52 @@ class DirectExtractionEngine:
page_num: Page number for logging page_num: Page number for logging
Returns: Returns:
List of dicts with covered text info: {'text', 'bbox', 'coverage'} List of dicts with covered text info: {'text', 'bbox', 'coverage', 'color_type'}
""" """
covered_words = [] covered_words = []
# Get all drawings and find white-filled rectangles # Get all drawings and find solid-filled rectangles
drawings = page.get_drawings() drawings = page.get_drawings()
white_rects = [] covering_rects = [] # List of (rect, color_type)
for d in drawings: for d in drawings:
fill_color = d.get('fill') fill_color = d.get('fill')
# Check for white fill (RGB all 1.0 or close to it) if fill_color and isinstance(fill_color, (tuple, list)) and len(fill_color) >= 3:
if fill_color:
if isinstance(fill_color, (tuple, list)) and len(fill_color) >= 3:
r, g, b = fill_color[:3] r, g, b = fill_color[:3]
# Allow slight tolerance for "almost white"
if r >= 0.95 and g >= 0.95 and b >= 0.95:
rect = d.get('rect') rect = d.get('rect')
if rect: if not rect:
white_rects.append(fitz.Rect(rect)) continue
if not white_rects: fitz_rect = fitz.Rect(rect)
# Skip very small rectangles (likely not covering blocks)
if fitz_rect.width < 5 or fitz_rect.height < 5:
continue
# Detect white rectangles (white-out / correction tape)
if r >= 0.95 and g >= 0.95 and b >= 0.95:
covering_rects.append((fitz_rect, 'white'))
# Detect black rectangles (redaction)
elif r <= 0.05 and g <= 0.05 and b <= 0.05:
covering_rects.append((fitz_rect, 'black'))
# Detect other solid colors (uniform fill that might be covering)
# Only consider if it's a solid color (low variance) and reasonably large
elif fitz_rect.width >= 20 and fitz_rect.height >= 10:
# Check if it's a saturated solid color (not gradient-like)
# For now, detect very dark or very light solid fills
avg_brightness = (r + g + b) / 3
if avg_brightness <= 0.1 or avg_brightness >= 0.9:
covering_rects.append((fitz_rect, 'solid'))
if not covering_rects:
return covered_words return covered_words
logger.debug(f"Page {page_num}: Found {len(white_rects)} white rectangles") # Log detected covering rectangles by type
white_count = sum(1 for _, t in covering_rects if t == 'white')
black_count = sum(1 for _, t in covering_rects if t == 'black')
solid_count = sum(1 for _, t in covering_rects if t == 'solid')
logger.debug(f"Page {page_num}: Found {len(covering_rects)} covering rectangles "
f"(white: {white_count}, black: {black_count}, other: {solid_count})")
# Get all text words with bounding boxes # Get all text words with bounding boxes
# words format: (x0, y0, x1, y1, word, block_no, line_no, word_no) # words format: (x0, y0, x1, y1, word, block_no, line_no, word_no)
@@ -1947,9 +1975,9 @@ class DirectExtractionEngine:
if word_area <= 0: if word_area <= 0:
continue continue
for white_rect in white_rects: for cover_rect, color_type in covering_rects:
# Calculate intersection # Calculate intersection
intersection = word_rect & white_rect intersection = word_rect & cover_rect
if intersection.is_empty: if intersection.is_empty:
continue continue
@@ -1961,7 +1989,8 @@ class DirectExtractionEngine:
covered_words.append({ covered_words.append({
'text': word_text, 'text': word_text,
'bbox': tuple(word_rect), 'bbox': tuple(word_rect),
'coverage': coverage_ratio 'coverage': coverage_ratio,
'color_type': color_type
}) })
break # Word is covered, no need to check other rects break # Word is covered, no need to check other rects