A New Denoising Method with Contourlet Transform

The contourlet coefficients of natural images are highly correlated across scales especially around boundaries, whereas contourlet coefficients of Gaussan white noise are hardly correlated. Therefore signal coefficients can be sorted out according to the different correlation characteristics between signals and noise. In this paper, a contourlet based denoising algorithm using interscale correlations is proposed combined with thresholding functions. Experimental results show that the proposed method outperforms the corresponding wavelet based method especially for the images containing many edges and fine textures.