Edge Detection from Candidate Boundaries and Qualitative Comparison of Results for Color Images

The local color cue based candidate boundary detection incorporating stationary Haar wavelet decomposition at various levels and thresholded proximity influence measures are used for detecting edges in color images in the proposed method. The non-homogeneous inter-tuples and non-uniform intra-tuple contributions of RGB tuples for conception of perceptual-edges are exploited in the edge detection method from candidate boundaries for color images. These edge detection results are presented here and qualitatively compared with those produced with leading image processing & desk-top-processing tools - Adobe Photoshop, MS Photo Editor and ACD Photo Editor for images of different resolutions, types and characteristics. The method results, incorporating different levels of stationary Haar wavelet decompositions are also compared, analyzed and presented. The qualitative comparison of results of thinned edges is also presented. The method results outperform others detecting perceptually significant edges, proving its suitability for edge features extraction, object detection & identification.