Research On Edge Preserving For Image Denoising Based On PDE

Image denoising based on partial diffusion equation becomes more and more popular in recent years due to it can coordinate the relationship between noise removing and edge preserving. With the classical P-M anisotropic diffusion model, although it can avoid edge blurring during the denoising at some extent, still has some inadequacy. In this paper, analyzes the principles and inadequacy based on P-M nonlinear anisotropic diffusion model, studied edge-persevering denoising algorithms based on partial differential equation thoroughly. The main subjects are as follows:Edge-preserving denoising algorithms based on the P-M nonlinear anisotropic diffusion model are studied. According to analyze the classical format of correlation coefficient, in order to overcome the lack of gradient threshold by hand, under the premise with not increase the variable, proposed a new optimized gradient threshold and adaptive method.It is an effective way for introduced the factor, which express the local statistics information for modified the classical P-M model. Therefore, a modified scheme that the local variation of pixel is incorporated in diffusion coefficient is proposed.The number of iterations for classical P-M model has a direct effect for denosing. In the short of fixed number of iteration and analyze the existing iteration stopping criterion, a new adaptive and no reference algorithm to determine the number of iterations based on local minimize correlation coefficient is proposed.Three proposed algorithms in this paper are all taken the experiments. Using standard test images and real images (two-photon microscopy images on olfactory bulb of mouse) for testing, the results have shown that the algorithms both have efficiency and practicality.