detectDocumentQuad static method

DocumentCorners detectDocumentQuad(Uint8List gray, int width, int height, { int candidateCount = 5, double minAreaRatio = 0.15, })

Detects the best document quadrilateral from a grayscale image.

This is the high-level API that chains: Canny edge detection → contour finding → quad extraction → corner sorting.

gray — Single-channel grayscale pixel buffer. width, height — Image dimensions. candidateCount — Number of top contour candidates to evaluate.

Returns the best DocumentCorners found, or a default inset quad if no valid document edges are detected.

Implementation

static DocumentCorners detectDocumentQuad(
  Uint8List gray,
  int width,
  int height, {
  int candidateCount = 5,
  double minAreaRatio = 0.15,
}) {
  // Stage 1: Canny edge detection
  final edges = CannyEdgeDetector.detect(
    gray,
    width,
    height,
    useAutoThreshold: true,
  );

  // Stage 2: Find contours
  final contours = findContours(edges, width, height, minContourLength: 40);

  // Stage 3: Evaluate top candidates for best quadrilateral
  DocumentCorners? bestQuad;
  double bestScore = 0;

  final evalCount = math.min(candidateCount, contours.length);
  for (int i = 0; i < evalCount; i++) {
    final quad = findBestQuadrilateral(
      contours[i],
      width,
      height,
      minAreaRatio: minAreaRatio,
    );

    if (quad != null) {
      final score = _scoreQuadrilateral(quad, width, height);
      if (score > bestScore) {
        bestScore = score;
        bestQuad = quad;
      }
    }
  }

  // Return best quad or default inset
  return bestQuad ??
      DocumentScannerService.detectDocumentEdges(
        Size(width.toDouble(), height.toDouble()),
      );
}