Color texture image retrieval based on Gaussian copula models of Gabor wavelets

Abstract

Color texture retrieval is a hot research area in image analysis. In this paper, we propose an efficient color texture retrieval method by using copula model based on Gabor wavelets. When Gabor wavelets are used to decompose color image, three types of dependence exist in the decomposed subbands of Gabor wavelets: color dependence, scale dependence and direction dependence. We analyze these dependencies and then capture them by using Gaussian copula function. Four copula schemes are developed, and accordingly four KLDs (Kullback-Leibler distances) of the copula schemes are introduced for color texture image retrieval. The evaluations of the proposed method are performed on several color texture databases including two large texture databases ALOT and STex. Experimental results demonstrate the proposed method has better performance than the state-of-the-art retrieval methods.