| Nowadays, the fact that noisy images not only affect the visual impression but also increase the difficulty of image analysis such that the technique of noise removing, which affects the quality of the images, becomes a key step in image pre-treatment.For not considering image feature, traditional image smoothing algorithms of isotropic diffusion method, such as mean filter, median filter and Gauss filter, get the denoised image with fuzzy edge. Such problem motivated the development of PDE-based image smoothing technology. It can preserve the edge of image while removing the image noise. Moreover, because of their good noise removing, feature and its directions' extraction, it is thought to be an adaptive smoothing method. Compared with isotropic diffusion method, anisotropic diffusion method is actually non-linear parabolic partial differential equations, estimating the speed of diffusion by using gradient to remove the noise and preserve feature well. Perona-Malik, the representative of such method, is already widely applied in the fields of marginal detection, image enhancement and image partition, etc.Image smoothing based on variation method, supported by strong foundations of mathematical theory and perfect resolving method of numerical calculation, is another part of partial differential equations' application. Plenty of models of such method have existed. The Total Variation model (TV) for image restoration proposed by Rudin, Osher, and Fatemi in 1992 is one of the classical methods among them. Because of its good edge-preserving, it is viewed as a better one to reconstruct the image while removing the noise. In this paper, based on the result that it can't reduce the salt and pepper noise well, we propose a new method to remove salt and pepper noise. Viewed salt and pepper noise as multiply model. Then use the basic thought of TV method to remove the salt and pepper noise. The experimental results show this method has good effect of removing salt and pepper noise. |
PDF Full Text Request