Research On Image Denoising Based On Contourlet Transform

Image denoising is a very important subject in signal processing. In recent years, contourlet-domain based image denoising algorithm becomes more and more popular in this field. As one of the most important methods among multiscale geometric analysis, contourlet transform has very good properties such as multidirection and anisotropy, as well as the properties of multiresolution and time-frequency localization inherited from wavelet transform. Besides, it has multidirection and anisotropy. Therefore it has promising applications to image denoising since it can capture significant information from a natural image.After studying the elementary concept of contourlet transform and obtainable image denoising methods, we focused on the correlation properties between contourlet coefficients both at the same scale and neighboring scales. Finally two new contourlet-domain image denoising algorithms are proposed based on the correlation properties of contourlet coefficients. The main contributions of this paper are summarized as follows.With the study of theory and realization of contourlet transform, we construct an adaptive threshold by means of intra-scale correlation property of contourlet coefficients. Therefore, we obtained an adaptive algorithm for image denoising based on correlation of neighboring coefficients at the same scale of contourlet transform. The proposed algorithm can get better denoising results than the methods available.With the statistical studies of the correlation of contourlet coefficients, another contourlet-domain image denoising algorithm based on inter-scale correlation of contourlet coefficients is proposed. This algorithm takes the correlation of inter-scale contourlet coefficients into consideration in the process of shrinkage. The new algorithm can not only remove noise thoroughly, but can effectively maintain the edges of images. It outperforms other denoising methods in terms of PSNR and visual qualities.