Study On The Method Of Image Denoising Based Onnonsubsampled Contourlet And Wavelet Transform

In this paper, we studied the methods of image denoising based on multi-scale decomposition named NonSubsample Contourlet Transform(NSCT) and Wavelet Transform(WT). For images decomposing , WT used two-dimensional tensor product has three directions and is isotropy and multiscale. NSCT is a flexible multiscale, multidirection, shift-invariant image decomposition. The aim is to develop computationally efficient, high-quality image denoising methods with the two transform.The research contents are as follows:①In this paper, the development of Multiscale Geometric Analysis(MGA) is depicted simply and the application of image denoising based on MGA and WT are introduced. The theory about the wavelet transform, the contourlet transform(CT) and the nonsubsampled contourlet transformare describesed, respectively.②The image denoising performances have been shown by the two different discussed tools and draw a conclusion that the smooth images denoising performance based on NSCT outperform the wavelet transform, the non-smooth images based on wavelet transform outperform NSCT.③In order to overcome the drawbacks of the single method discussed above, a novel algorithm for denoising images'region segmentation is proposed based on features of the denoising images'deviation. And using the segmenting algorithm, a series of the denoising methods with region segmentation are proposed, such as image denoising based on region segmentation with wavelet transform, with contourlet and NSCT.④Gauss Scale Mixture(GSM) Modle is introduced detailed. The paper characterizes the coefficients of NSCT by GSM, and a denoising algorithm based on GSM and Bayes Least Squares Estimator(BLSE) is proposed.In order to demonstrate the effectiveness of our proposed method, the algorithm has been used to image denoising. The experimental results indicate that the proposed approach can avoid the introduction of artifacts and can signigicantly outperform the traditional wavelet-based,linear shift invariant wavelet-based, contourlet-based and nonsubsampled contourlet-based image denosing methods singly using in terms of both visual quality and objective evaluation criteria.