breakpilot-lehrer/klausur-service/backend/ocr/detect/box_detect.py

"""
Embedded box detection and page zone splitting for the CV vocabulary pipeline.

Detects boxes (grammar tips, exercises, etc.) that span the page width and
interrupt the normal column layout. Splits the page into vertical zones so
that column detection can run independently per zone.

Two-stage algorithm (both run, results merged):
  1. Morphological line detection — finds bordered boxes via horizontal lines.
  2. Background shading detection — finds shaded/colored boxes via median-blur
     background analysis.  Works for colored (blue, green) and grayscale
     (gray shading on B/W scans) boxes.

Lizenz: Apache 2.0 (kommerziell nutzbar)
DATENSCHUTZ: Alle Verarbeitung erfolgt lokal.
"""

import logging
from typing import List, Optional, Tuple

import cv2
import numpy as np

from ..types import DetectedBox, PageZone

logger = logging.getLogger(__name__)

__all__ = [
    "detect_boxes",
    "split_page_into_zones",
]


# ---------------------------------------------------------------------------
# Stage 1: Morphological line detection
# ---------------------------------------------------------------------------

def _detect_boxes_by_lines(
    gray: np.ndarray,
    content_x: int,
    content_w: int,
    content_y: int,
    content_h: int,
) -> List[DetectedBox]:
    """Find boxes defined by pairs of long horizontal border lines.

    Args:
        gray: Grayscale image (full page).
        content_x, content_w: Horizontal content bounds.
        content_y, content_h: Vertical content bounds.

    Returns:
        List of DetectedBox for each detected bordered box.
    """
    h, w = gray.shape[:2]

    # Binarize: dark pixels → white on black background
    _, binary = cv2.threshold(gray, 180, 255, cv2.THRESH_BINARY_INV)

    # Horizontal morphology kernel — at least 50% of content width
    kernel_w = max(50, content_w // 2)
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_w, 1))
    lines_img = cv2.morphologyEx(binary, cv2.MORPH_OPEN, kernel)

    # Horizontal projection: count line pixels per row
    h_proj = np.sum(lines_img[:, content_x:content_x + content_w] > 0, axis=1)
    line_threshold = content_w * 0.30

    # Group consecutive rows with enough line pixels into line segments
    line_segments: List[Tuple[int, int]] = []  # (y_start, y_end)
    seg_start: Optional[int] = None
    for y in range(h):
        if h_proj[y] >= line_threshold:
            if seg_start is None:
                seg_start = y
        else:
            if seg_start is not None:
                line_segments.append((seg_start, y))
                seg_start = None
    if seg_start is not None:
        line_segments.append((seg_start, h))

    if len(line_segments) < 2:
        return []

    # Pair lines into boxes: top-line + bottom-line
    # Minimum box height: 30px.  Maximum: 70% of content height.
    min_box_h = 30
    max_box_h = int(content_h * 0.70)

    boxes: List[DetectedBox] = []
    used = set()
    for i, (top_start, top_end) in enumerate(line_segments):
        if i in used:
            continue
        for j in range(i + 1, len(line_segments)):
            if j in used:
                continue
            bot_start, bot_end = line_segments[j]
            box_y = top_start
            box_h = bot_end - top_start
            if box_h < min_box_h or box_h > max_box_h:
                continue

            # Estimate border thickness from line segment heights
            border_top = top_end - top_start
            border_bot = bot_end - bot_start

            box = DetectedBox(
                x=content_x,
                y=box_y,
                width=content_w,
                height=box_h,
                confidence=0.8,
                border_thickness=max(border_top, border_bot),
            )
            boxes.append(box)
            used.add(i)
            used.add(j)
            break  # move to next top-line candidate

    return boxes


# ---------------------------------------------------------------------------
# Stage 2: Background shading detection (color + grayscale)
# ---------------------------------------------------------------------------

def _detect_boxes_by_shading(
    img_bgr: np.ndarray,
    content_x: int,
    content_w: int,
    content_y: int,
    content_h: int,
) -> List[DetectedBox]:
    """Find boxes with shaded/colored background (no visible border lines).

    Uses heavy median blur to remove text and reveal the underlying background.
    Then detects rectangular regions where the background differs from white.
    Works for both colored boxes (blue, green) and grayscale shading (gray on
    B/W scans).

    Args:
        img_bgr: BGR color image (full page).
        content_x, content_w: Horizontal content bounds.
        content_y, content_h: Vertical content bounds.

    Returns:
        List of DetectedBox for each detected shaded box.
    """
    h, w = img_bgr.shape[:2]

    # --- Heavy median blur removes text strokes, keeps background ---
    blur_size = 31  # large kernel to wipe out text
    blurred = cv2.medianBlur(img_bgr, blur_size)
    blur_gray = cv2.cvtColor(blurred, cv2.COLOR_BGR2GRAY)
    blur_hsv = cv2.cvtColor(blurred, cv2.COLOR_BGR2HSV)

    # Estimate page background from top-left / top-right corners
    corner_size = max(20, min(h // 10, w // 10))
    corners = np.concatenate([
        blur_gray[:corner_size, :corner_size].ravel(),
        blur_gray[:corner_size, -corner_size:].ravel(),
    ])
    page_bg = float(np.median(corners))

    # Two masks: grayscale shading + color saturation
    # Grayscale: regions noticeably darker than the page background
    shade_thresh = max(page_bg - 30, 150)
    gray_mask = (blur_gray < shade_thresh).astype(np.uint8) * 255

    # Color: regions with noticeable saturation (blue/green/etc. boxes)
    sat_mask = (blur_hsv[:, :, 1] > 20).astype(np.uint8) * 255

    combined = cv2.bitwise_or(gray_mask, sat_mask)

    # Morphological cleanup: close gaps, remove small noise
    kernel_close = cv2.getStructuringElement(cv2.MORPH_RECT, (25, 10))
    combined = cv2.morphologyEx(combined, cv2.MORPH_CLOSE, kernel_close)
    kernel_open = cv2.getStructuringElement(cv2.MORPH_RECT, (10, 5))
    combined = cv2.morphologyEx(combined, cv2.MORPH_OPEN, kernel_open)

    contours, _ = cv2.findContours(combined, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    # Size thresholds: smaller boxes allowed (e.g. "German leihen" ~30% width)
    min_area = content_w * 30  # at least 30px tall at full width
    min_box_h = 25
    max_box_h = int(content_h * 0.70)
    min_width_ratio = 0.25  # boxes can be ~25% of content width

    boxes: List[DetectedBox] = []
    for cnt in contours:
        area = cv2.contourArea(cnt)
        if area < min_area:
            continue

        bx, by, bw, bh = cv2.boundingRect(cnt)

        # Width filter
        if bw < content_w * min_width_ratio:
            continue

        # Height filter
        if bh < min_box_h or bh > max_box_h:
            continue

        # Rectangularity check: area / bounding-rect area > 0.6
        rect_area = bw * bh
        if rect_area > 0 and area / rect_area < 0.5:
            continue

        # Verify that the background inside this region is actually shaded
        roi_gray = blur_gray[by:by + bh, bx:bx + bw]
        roi_hsv = blur_hsv[by:by + bh, bx:bx + bw]
        if roi_gray.size == 0:
            continue

        median_val = float(np.median(roi_gray))
        median_sat = float(np.median(roi_hsv[:, :, 1]))

        # Must be noticeably different from page background
        is_shaded = median_val < (page_bg - 15)
        is_colored = median_sat > 15

        if not is_shaded and not is_colored:
            continue

        conf = 0.7 if is_colored else 0.6

        boxes.append(DetectedBox(
            x=bx,
            y=by,
            width=bw,
            height=bh,
            confidence=conf,
            border_thickness=0,
        ))

    return boxes


# ---------------------------------------------------------------------------
# Validation
# ---------------------------------------------------------------------------

def _validate_box(
    box: DetectedBox,
    gray: np.ndarray,
    content_w: int,
    content_h: int,
    median_row_gap: int,
) -> bool:
    """Validate that a detected box is genuine (not a table-row separator etc.)."""
    # Must span > 25% of content width (lowered from 60% to allow smaller boxes)
    if box.width < content_w * 0.25:
        return False

    # Height constraints
    if box.height < 25 or box.height > content_h * 0.70:
        return False

    # Must not be confused with a table-row separator:
    # real boxes are at least 3x the median row gap
    if median_row_gap > 0 and box.height < median_row_gap * 3:
        return False

    # Must contain some text (ink density check)
    h, w = gray.shape[:2]
    y1 = max(0, box.y)
    y2 = min(h, box.y + box.height)
    x1 = max(0, box.x)
    x2 = min(w, box.x + box.width)
    roi = gray[y1:y2, x1:x2]
    if roi.size == 0:
        return False
    ink_ratio = np.sum(roi < 128) / roi.size
    if ink_ratio < 0.002:  # nearly empty → not a real content box
        return False

    return True


# ---------------------------------------------------------------------------
# Public API: detect_boxes
# ---------------------------------------------------------------------------

def _merge_overlapping_boxes(boxes: List[DetectedBox]) -> List[DetectedBox]:
    """Merge boxes that overlap significantly (IoU > 0.3 or one contains the other).

    When two boxes overlap, keep the one with higher confidence (or the larger
    one if confidences are equal).
    """
    if len(boxes) <= 1:
        return boxes

    # Sort by area descending so larger boxes are processed first
    boxes = sorted(boxes, key=lambda b: b.width * b.height, reverse=True)
    keep = [True] * len(boxes)

    for i in range(len(boxes)):
        if not keep[i]:
            continue
        bi = boxes[i]
        for j in range(i + 1, len(boxes)):
            if not keep[j]:
                continue
            bj = boxes[j]

            # Compute overlap
            x1 = max(bi.x, bj.x)
            y1 = max(bi.y, bj.y)
            x2 = min(bi.x + bi.width, bj.x + bj.width)
            y2 = min(bi.y + bi.height, bj.y + bj.height)

            if x2 <= x1 or y2 <= y1:
                continue  # no overlap

            inter = (x2 - x1) * (y2 - y1)
            area_i = bi.width * bi.height
            area_j = bj.width * bj.height
            smaller_area = min(area_i, area_j)

            # If overlap covers > 50% of the smaller box, merge (drop the weaker)
            if smaller_area > 0 and inter / smaller_area > 0.50:
                # Keep the one with higher confidence; if equal, keep larger
                if bj.confidence > bi.confidence:
                    keep[i] = False
                    break
                else:
                    keep[j] = False

    return [b for b, k in zip(boxes, keep) if k]


def detect_boxes(
    img_bgr: np.ndarray,
    content_x: int,
    content_w: int,
    content_y: int,
    content_h: int,
    median_row_gap: int = 0,
) -> List[DetectedBox]:
    """Detect embedded boxes on a page image.

    Runs BOTH line-based and shading-based detection, then merges and
    deduplicates results.

    Args:
        img_bgr: BGR color image (full page or cropped).
        content_x, content_w: Horizontal content bounds.
        content_y, content_h: Vertical content bounds.
        median_row_gap: Median row gap height (for filtering out table separators).

    Returns:
        List of validated DetectedBox instances, sorted by y position.
    """
    gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

    # Stage 1: Line-based detection (bordered boxes)
    line_boxes = _detect_boxes_by_lines(gray, content_x, content_w, content_y, content_h)

    # Stage 2: Shading-based detection (colored/gray background boxes)
    shade_boxes = _detect_boxes_by_shading(img_bgr, content_x, content_w, content_y, content_h)

    logger.debug("BoxDetect: %d line-based, %d shading-based candidates",
                 len(line_boxes), len(shade_boxes))

    # Combine and deduplicate
    all_boxes = line_boxes + shade_boxes
    merged = _merge_overlapping_boxes(all_boxes)

    # Validate
    validated = [b for b in merged if _validate_box(b, gray, content_w, content_h, median_row_gap)]

    # Sort top to bottom
    validated.sort(key=lambda b: b.y)

    if validated:
        logger.info("BoxDetect: %d box(es) detected (line=%d, shade=%d, merged=%d)",
                     len(validated), len(line_boxes), len(shade_boxes), len(merged))
    else:
        logger.debug("BoxDetect: no boxes detected")

    return validated


# ---------------------------------------------------------------------------
# Zone Splitting
# ---------------------------------------------------------------------------

def split_page_into_zones(
    content_x: int,
    content_y: int,
    content_w: int,
    content_h: int,
    boxes: List[DetectedBox],
    min_zone_height: int = 40,
) -> List[PageZone]:
    """Split a page into vertical zones based on detected boxes.

    Regions above, between, and below boxes become 'content' zones;
    box regions become 'box' zones.

    Args:
        content_x, content_y, content_w, content_h: Content area bounds.
        boxes: Detected boxes, sorted by y position.
        min_zone_height: Minimum height for a content zone to be kept.

    Returns:
        List of PageZone, ordered top to bottom.
    """
    if not boxes:
        # Single zone: entire content area
        return [PageZone(
            index=0,
            zone_type='content',
            y=content_y,
            height=content_h,
            x=content_x,
            width=content_w,
        )]

    zones: List[PageZone] = []
    zone_idx = 0
    cursor_y = content_y
    content_bottom = content_y + content_h

    for box in boxes:
        # Content zone above this box
        gap_above = box.y - cursor_y
        if gap_above >= min_zone_height:
            zones.append(PageZone(
                index=zone_idx,
                zone_type='content',
                y=cursor_y,
                height=gap_above,
                x=content_x,
                width=content_w,
            ))
            zone_idx += 1

        # Box zone
        zones.append(PageZone(
            index=zone_idx,
            zone_type='box',
            y=box.y,
            height=box.height,
            x=box.x,
            width=box.width,
            box=box,
        ))
        zone_idx += 1

        cursor_y = box.y + box.height

    # Content zone below last box
    remaining = content_bottom - cursor_y
    if remaining >= min_zone_height:
        zones.append(PageZone(
            index=zone_idx,
            zone_type='content',
            y=cursor_y,
            height=remaining,
            x=content_x,
            width=content_w,
        ))

    logger.info(f"ZoneSplit: {len(zones)} zones from {len(boxes)} box(es): "
                f"{[z.zone_type for z in zones]}")

    return zones