Texture feature extraction and description using fuzzy set of main dominant directions of variable scales in content-based medical image retrieval

Abstract

The paper presents a new method for texture feature extraction and description. Gabor wavelet transform, statistical method and fuzzy logic are used together to compute the fuzzy set of main dominant directions of variable scales of medical images of the same kind and the membership degree of dominance of each element in the fuzzy set. Texture feature vector of each image is computed based on the fuzzy set. Moreover, the dominance of each element in the fuzzy set is introduced into the similarity measure as weight. Experiments show that the method proposed in this paper reduces the dimension of texture feature vector and alleviates heavy computation while keeping and even improving the retrieval performance.