TEXTURE IMAGE RETRIEVAL USING NONSUBSAMPLED CONTOURLET TRANSFORM

Texture retrieval is a significant branch of content based image retrieval (CBIR). In this paper, the nonsubsampled contourlet transform has been used for texture image retrieval. This transform is built upon nonsubsampled pyramid and nonsubsampled directional filter banks and provides a shift-invariant, directional, multi-resolution image representation. The statistical characteristics of these directional subbands coefficients have been used as the feature vector for texture image classification. The performance of the proposed method is compared to other texture image retrieval methods such as Gabor filter, Contourlet and Complex Directional Filter Bank (CDFB) and the experimental results show that our proposed method improves retrieval rate and outperforms other approaches in this concept.