Research On Multi-shape Estimation Based On NLM Image Denoising Algorithm

There is a lot of similar structure redundancy information in natural images by constructing a neighborhood block between the structural characteristics of the pixel vectors to obtain better reflect the effect of the non-local means(NLM)image denoising algorithm uses.However,since the NLM algorithm in the image of high contrast edges coincide rational calculation of similarity weight is poor,and therefore easy to introduce at the edge of shock "artifacts" phenomenon.we proposed a new multi-shape estimation for this problem nonlocal mean denoising algorithm.The algorithm used stationary wavelet transform over complete noise characteristic multiscale image decomposition to calculate the pre-denoised image.Computation of the gradient pixel similarity factor by constructing gradient vector.We using general shapes instead of square patches.A fast FFT-based algorithm is proposed to compute the NLM with arbitray shapes.And then,the local combination of the different shapes relies on Stein's Unbiased Risk Estimate.Finally,to improve the robustness of this local aggregation,we perform an anistropic diffusion of the risk estimate using a modified P-Mequation.Experiment results are given to demonstrate the proposed algorithm perform better in both high contrast eages and PSNR/SSIM values.when the noise standard deviation is large enough,the proposed approach has effectiveness and robustness.