Research On The Algorithm Of Image Denoising Based On Anisotropic Diffusion And Wavelet Transform

Image processing and analysis is a major field of information science and engineering. During the forming, transferring and memorizing of the digital images, the images are often corrupted by different kinds of noises. Therefore, the quality of image became worse.Image denoising becomes one of the most important steps in image processing.Wavelet transform and anisotropic diffusion are two important methods of image denoising. In this thesis, firstly, the image denoising based on wavelet transform and anisotropic diffusion are researched separately. Secondly, the correlation between wavelet transform and anisotropic diffusion in image denoising is discussed. Thirdly, the relationship between nonlinear diffusion and rotationally invariant wavelet shrinkage is analyzed. Finally, the equivalence between nonlinear diffusion and rotationally invariant wavelet shrinkage is proved.As to P-M method, it is always a problem of finding whether it is noise or edge of the image. F.Catte has improved the method by convolving the image with a Gaussian prior to differentiation which has been inherited in almost all methods proposed later. But it also has its own problem because it is difficult to choose the Gaussian Kernel. In this thesis, an innovative nolinear wavelet shrinkage method used to replace the Gaussian filter is inspired by the equivalence and is used in Catte and coherence enhancing diffusion (CED).