CORNER Method for Edge Detection - Corner Detection in Intensity Images

A practical edge detection method is CORNER, which identifies corner points in intensity images. The CORNER algorithm works through the following implementation steps:

1. First, apply the Canny edge detector to the grayscale image using functions like edge(I, 'Canny') to obtain a binary edge map with optimal thresholding for noise reduction.

2. Extract edge contours from the edge map using contour tracing algorithms (e.g., bwboundaries in MATLAB) and fill gaps in the contours through morphological operations or interpolation techniques.

3. Calculate curvature for each contour at low scale using derivative filters (e.g., Sobel or Gaussian derivatives) to preserve genuine corner points while suppressing noise.

4. All local curvature maxima are considered corner candidates, with subsequent filtering to eliminate round corners and false detections caused by boundary noise and fine details using threshold-based validation.

5. If endpoints of linear curves aren't proximate to previously detected corners, they're added as corner points through distance calculation and proximity checks.

This method provides comprehensive corner detection in images. By filling gaps in edge contours, it enhances detection accuracy through robust contour completion algorithms. Furthermore, by eliminating unqualified round corners and false detections via curvature analysis and validation steps, it significantly improves result reliability.