A Comparative Performance Analysis of CBIR using Wavelet and Contourlet Transform

Abstract

In the literature various algorithms for content based image retrieval (CBIR) were proposed on different variety of databases such as texture, face, and medical images. It is observed that the CBIR performance strongly depends on effective extraction of image features such as color, texture, and shape. In the proposed research work, contourlet transform based CBIR system is proposed for texture images. Similar to wavelets, texture images can be characterized via marginal distributions of subband coefficients of recently introduced contourlets. In addition, contourlets have two new properties namely directionality and anisotropy which makes them powerful in efficient representation of natural texture images. Hence, in the proposed research work, we have analyzed and compared the performance of both wavelet and contourlet schemes for content based texture image retrieval. It is observed that contourlet based method outperforms the wavelet method for highly texture images in terms of retrieval efficiency.