Sparse representation of natural image based on Contourlet overcomplete dictionary

In this paper, textural features of natural image are researched from the perspective of highly nonlinear approximation theory, in view of the characteristic that a contour of natural image is composed of piecewise regular geometrical curves of the image plane. According to nonlinear approximation theory and multi-scale geometric analysis method, a dictionary based on Contourlet basis function is also proposed in this paper. This dictionary approximates texture area of image by using Orthogonal Matching Pursuit (OMP) method. Experimental results show that Peak Signal to Noise Ratio (PSNR) and Sparsity Ratio (SR) of nonlinear approximated images can be improved effectively by using the proposed dictionary.