A Multiscale Gradient Fusion Method for Edge Detection in Color Images Utilizing the CBM3D Filter

This paper presents a novel color edge detection approach that integrates collaborative filtering with multiscale gradient fusion. After transforming RGB images into the XYZ color space, a block-matching and 3D (BM3D) filter is applied to enhance sparse representations and reduce noise. To address detail loss in single-scale edge detection, anisotropic Gaussian directional derivatives are calculated at multiple scales, then fused with vector gradients to form a robust edge strength map. Normalization, non-maximum suppression, double-threshold selection, and morphological refinement are used to finalize the edge contours. Experimental results on standard datasets show strong noise resistance and high-quality edges, with PR curves, AUC, PSNR, MSE, and FOM metrics confirming that this method outperforms Color Sobel, Color Canny, SE, and Color AGDD.