Anisotropic Diffusion Model With Fidelity Term

Image denoising is a very impotant field in Image processing.On the one hand,itcan provede more precise information for subsequent processing,such as edgedetectionand objectrecognition.On the other hand,the study of image denoising method cancontribute to the solution of other problems of image andanalysis,such asimagerestoration,super-resolution reconstruction.Currently,there are many methodsapplied to the image denoising[1-6],for example median filtering,anisotropic diffusionmethod.In recent years.The Partial Differential Equation has a maturity and a widerange of applications in physics, especially effective for the diffusion model.Anisotropic diffusion model is a typical method of partial differential equations forimage denoising.Using this model can not only better remove noise.And can better keepimage edge features.Perona and Malilik first proposed an anisotropic diffusionequation[7](P-M model)for image denoising.The TV model[8]is also an anisotropicdiffusion equation.Based on anisotropic diffusion equation.This paper made the following research:The fidelity term have great significance in image denoising.Because we requirenot only remove the noise but also has the effect of fidelity.But theTVmodel’s globalfidelity term also retained the noise.So based on this idea, we proposed a selectivefidelity term,and applied to the diffusion model.As the P-M model diffusion coefficientis only related with the gradient values,andwithout considering the gray.While the details of the image gradient values alsogenerally small,thus making the details are smoothed.To address this issue,many modelswere proposed[9,10].In this paper we foucs the Anisotropic diffusion combined with localentropy(Zhao model)[10]. We perform this selective fidelity term on the Zhao model, andpropose a novel noise removal model. Compared with the Zhao model, the proposedmodel not only can preserve the edges and texture better, but also can more effectivelyremove noise on experimental results.