Edge Information in Color Images Based on Histograms of Differences

摘要

A new measure of edge information for color images based on cumulative histograms of absolute color differences is proposed. A multispectral version of the Symmetric Nearest Neighbor filter for edge-preserving smoothing and methods for image segmentation and edge detection are developed based on this measure. Experimental results show that the performance of the new algorithms is very good. The analysis of color images has received little attention in computer vision research, in spite of the facts that color plays an important role in human vision and that color should also provide much useful information for many image analysis applications.