Reduced feature texture retrieval using contourlet decomposition of luminance image component

In this paper, a texture retrieval system based on directional hidden Markov model (HMM) in the contourlet domain is described. Through a contourlet transform, a directional multiscale transformation, the luminance component of an image can be decomposed into a set of directional subbands with texture details captured in different orientations at various scales. By exploiting in-band spatial dependencies, the distribution of the coefficients in each subband, which is modeled as a Gaussian mixture, is estimated using a directional hidden Markov model. We compare retrieval systems on the basis of retrieval rate and find that the proposed HMM exploiting in-band luminance dependencies provides reasonable results with much fewer features.