Research And Analysis Of The Image Denoising Algorithm Based On Partial Differential Equation

With the rapid development of digital and information technology, digital image has become an important source of getting information. In engineering practice, because of the imperfect of equipments and uncontrollable of the external environment, images are difficult to avoid the interference of noise in the process of collection, transport and storage. Noise can blur the image, and make the interested target difficult to identify, which seriously affect the next-process and analysis of image, therefore the noise removing of the image is especially important in the field of image processing. However, the traditional image denoising algorithms will blur the structure of the image, while filtering the noise, which is the problem of image denoising models, the contradiction between the reduction of the noise and the reservation of the structure characteristics. Image denoising algorithms based on partial differential equation can adopt different smoothing strategies according to the structure characteristics of the image, and can achieve more satisfactory effect in image denoising. The image denoising algorithm based on partial differential equation is studied and analyzed in the thesis. The main work is organized as follows:1. Chao-Tsai model uses the gradient information and gray variance of image to detect edge, and improves anisotropic diffusion model. Because difference curvature operator do better than gray-scale variance in the distinction between weak edges and noise, the difference curvature operator was introduced into the anisotropic diffusion model instead of gray-level variance. The article constructed a new adaptive diffusion coefficient function, and put forward a new model of noise reduction. Results of the experiment verify the superiority of the proposed algorithm for noise reduction.2. The thesis analyzed the diffusion mechanism of the anisotropic diffusion, andproposed a new model to overcome the speckle noise problem in Chao-Tsai model. As the rank-ordered absolute differences operator can better distinguish the weak edge and noise, the improved noise reduction model took advantage of the operator, and introduced it into the Chao-Tsai model. The experimental results show that, compared with the classical denoising model, the processed image of the new model is more clear, and the new model has better denoising effect. Furthermore, the new model can effectively deal with mixed noise including salt and pepper noise and gaussian noise, and it has strong applicability.3. Because most algorithms of noise reduction based on partial differential equation only compare the center pixel with a single one, that means these algorithms have one-sidedness in the process of denoising. Therefore, the thesis introduced the notion of the similarity into the anisotropic diffusion model based on the variable exponent, and put forward a new model of noise reduction. The model not only can select different noise reduction models in different regions of the image, but also can make full use of pixel information in the patch of the image,which make the noise reduction effect is improved obviously.4. MATLAB software was used to verify the proposed three kinds of noise reduction algorithms based on partial differential equation. By adopting the method of subjective evaluation and objective evaluation method, the thesis analyzed the denoising image obtained from the experiment respectively. Visual effect and objective evaluation parameters show that the proposed three kinds of improved algorithm has better denoising effect, and has higher application value.