The Research For Images Denoising Based On Improved Algorithm Of Anisotropic Diffusion

Image denoising is an important preliminary work in image processing, which involves removing noise and keeping the edge details.This contradiction has been plagued by image processing researchers. The anisotropic diffusion algorithm is compared with other algorithms,which has more advantage in the aspect of balancing noise removal with edge details.Therefore, the anisotropic diffusion algorithm has become one of research hotspots in the field of image denoising. After years of research and development,anisotropic diffusion algorithm has made many research results. However, in the aspect of image noise removal and remaining the edge details, there are still insufficient. To solve this problem, we propose two improved anisotropic diffusion algorithms, mainly as follows:(1) This paper gives a discussion on the denoising algorithm of anisotropic diffusion of images. Summarizing the existing researches, and on the basis of the smoothness of the spread function and the grade variation of the images, the writer builds a relationship between grade of the regional image and the spread function in order to realize adaptive selection of the spread function, and provide an improved anisotropic spread function model. It can solve the problem of many isolated noise points that the traditional PM model filter has. Meanwhile the edge-preserving of the remote sensing image details can achieve better result. The experiment shows that the proposed spread model in this paper can smooth and preserve the image edge well;(2) In view of the problem that anisotropic diffusion algorithm easily blurred image details and edges and denoising of incomplete, combining the LCC diffusion function with ECU diffusion function by using local variance information adjustment parameters and comprehensively utilization of local information described by local variance and gradient information, this paper proposes an anisotropic diffusion algorithm combined with local variance information. The model not only depends on the local variance and gradient information at the same time, but also in different area can timely adjusts diffusion velocity of LCC diffusion function and ECU diffusion function according to the local variance information adjustment parameters, making full use of the advantage of LCC diffusion function and ECU diffusion function. The experimental results show that compared withseveral other classical algorithms, the model not only can effectively remove the noise and preserving the image edge, but also to keep the details of image also has a better effect.(3) The above two kinds of image denoising models are tested by MATLAB. The subjective evaluation and objective evaluation are made to the denoising image,and we can draw the conclusion that the results of theoretical analysis are consistent with the experimental conclusion and that it not only can effectively remove noise, but also has good results in the aspect of maintaining image weak edge and details.