Use of contourlets for image retrieval

Abstract

We propose a simple technique for content-based image retrieval (CBIR) using contourlets that possess not only the multi-resolution time frequency localization properties of wavelets but are also good at capturing the directional information. Statistical features extracted from contourlet coefficients on each subband of the decomposed image, capturing the textural properties, are combined with the color feature while doing the CBIR. The retrieval performance on a large image database shows that the proposed technique provides a better accuracy compared to other multi-resolution based approaches. © Springer-Verlag Berlin Heidelberg 2005.