A Novel Decomposition Scheme for Image De-Noising

In this paper we introduce a novel decomposition scheme for image de-noising. The processed image is decomposed in several components following a tree structure similar to wavelet decomposition. The main difference rely on the fact that image components are obtained at the output of some adaptive filters while wavelet decomposition uses filters with fixed frequency response. Moreover, in our approach the resolution of the different components is the same. The components are filtered separately and recombined to obtain the output image. The proposed de-noising scheme shows improved performances and simpler implementation compared to other approaches based on wavelet, contourlet and other image transforms.