Image coding via sparse contourlet representation

When contourlet coefficients are directly coded, the benefits from directional representation are discounted by the redundancy of the transform. In this paper, we propose a contourlet-based image coding scheme under the sparse representation framework, in which contourlet coefficients are sparsified by the iterative thresholding method before compression. Dependency analysis is performed to reveal correlations among sparsified contourlet coefficients. Considering the characteristics of coefficient correlations, the obtained coefficients are coded with an intra-subband coder. Experimental results show that the coding performances of the contourlet-based schemes are significantly boosted via sparse representations.