Color texture retrieval based on fuzzy discriminate analysis

In this paper, a fuzzy discriminate analysis is proposed to apply for content based color texture retrieval. The feature of color texture can be characterized with color, tint, structure, pattern, weak regularity and randomness. The wavelet packet can be used to extract texture features. However, the dimension of feature vectors becomes very large, which is not suitable for texture retrieval system applications. In order to reduce the dimension of feature vectors, a fuzzy discriminate analysis is introduced to extract the principal basis of the wavelet moment. The extracted feature vector is robust for rotation and shift of texture structure and pattern. By unifying with the fuzzy color histogram, the fuzzy based measure for color texture retrieval is obtained. It is shown that the desired texture is effectively retrieved by the proposed method in simulations.