Texture analysis and clustering using the Wigner distribution

Abstract

Clustering (or grouping), defined as the taxonomy of objects (image regions) into categories based on the similarity of their characteristics, is examined. Both in computer and human vision, the use of joint spatial/spatial-frequency (s/sf) representations in this process are suggested. The Wigner and pseudo-Wigner distributions, examples of s/sf representations, are discussed. A comparison is made between the Wigner distribution, the spectrogram, the difference-of-Gaussians representation, and the Gabor representation. Experimental results in the areas of texture segmentation and Gestalt grouping using the Wigner distribution are presented, proving the feasibility of such an approach. >