Restructure: Move ocr_pipeline + labeling + crop into ocr/ package

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

klausur-service/backend/scan_quality.py

@@ -1,102 +1,4 @@
-"""
-Scan Quality Assessment — Measures image quality before OCR.
-
-Computes blur score, contrast score, and an overall quality rating.
-Used to gate enhancement steps and warn users about degraded scans.
-
-All operations use OpenCV (Apache-2.0), no additional dependencies.
-"""
-
-import logging
-from dataclasses import dataclass, asdict
-from typing import Dict, Any
-
-import cv2
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-# Thresholds (empirically tuned on textbook scans)
-BLUR_THRESHOLD = 100.0       # Laplacian variance below this = blurry
-CONTRAST_THRESHOLD = 40.0    # Grayscale stddev below this = low contrast
-CONFIDENCE_GOOD = 40         # OCR min confidence for good scans
-CONFIDENCE_DEGRADED = 30     # OCR min confidence for degraded scans
-
-
-@dataclass
-class ScanQualityReport:
-    """Result of scan quality assessment."""
-    blur_score: float         # Laplacian variance (higher = sharper)
-    contrast_score: float     # Grayscale std deviation (higher = more contrast)
-    brightness: float         # Mean grayscale value (0-255)
-    is_blurry: bool
-    is_low_contrast: bool
-    is_degraded: bool         # True if any quality issue detected
-    quality_pct: int          # 0-100 overall quality estimate
-    recommended_min_conf: int # Recommended OCR confidence threshold
-
-    def to_dict(self) -> Dict[str, Any]:
-        return asdict(self)
-
-
-def score_scan_quality(img_bgr: np.ndarray) -> ScanQualityReport:
-    """
-    Assess the quality of a scanned image.
-
-    Uses:
-    - Laplacian variance for blur detection
-    - Grayscale standard deviation for contrast
-    - Mean brightness for exposure assessment
-
-    Args:
-        img_bgr: BGR image (numpy array from OpenCV)
-
-    Returns:
-        ScanQualityReport with scores and recommendations
-    """
-    gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)
-
-    # Blur detection: Laplacian variance
-    # Higher = sharper edges = better quality
-    laplacian = cv2.Laplacian(gray, cv2.CV_64F)
-    blur_score = float(laplacian.var())
-
-    # Contrast: standard deviation of grayscale
-    contrast_score = float(np.std(gray))
-
-    # Brightness: mean grayscale
-    brightness = float(np.mean(gray))
-
-    # Quality flags
-    is_blurry = blur_score < BLUR_THRESHOLD
-    is_low_contrast = contrast_score < CONTRAST_THRESHOLD
-    is_degraded = is_blurry or is_low_contrast
-
-    # Overall quality percentage (simple weighted combination)
-    blur_pct = min(100, blur_score / BLUR_THRESHOLD * 50)
-    contrast_pct = min(100, contrast_score / CONTRAST_THRESHOLD * 50)
-    quality_pct = int(min(100, blur_pct + contrast_pct))
-
-    # Recommended confidence threshold
-    recommended_min_conf = CONFIDENCE_DEGRADED if is_degraded else CONFIDENCE_GOOD
-
-    report = ScanQualityReport(
-        blur_score=round(blur_score, 1),
-        contrast_score=round(contrast_score, 1),
-        brightness=round(brightness, 1),
-        is_blurry=is_blurry,
-        is_low_contrast=is_low_contrast,
-        is_degraded=is_degraded,
-        quality_pct=quality_pct,
-        recommended_min_conf=recommended_min_conf,
-    )
-
-    logger.info(
-        f"Scan quality: blur={report.blur_score} "
-        f"contrast={report.contrast_score} "
-        f"quality={report.quality_pct}% "
-        f"degraded={report.is_degraded} "
-        f"min_conf={report.recommended_min_conf}"
-    )
-
-    return report
+# Backward-compat shim -- module moved to ocr/pipeline/scan_quality.py
+import importlib as _importlib
+import sys as _sys
+_sys.modules[__name__] = _importlib.import_module("ocr.pipeline.scan_quality")