A Partial Differential ANRDPM Image Denoising Model Based on A New Anti-Noise Coefficient and Reverse Diffusion Idea

Abstract

 In order to overcome the problem of insufficient expression of fine texture when gradient mode is used as image feature extraction operator in traditional PM model, which leads to excessive diffusion in these fine texture regions and texture ambiguity, this paper proposes ANRDPM noise reduction model based on the new anti-noise coefficient and reverse diffusion concept. In this model, the meter gradient operator is used as the image feature extractor to solve the shortage of the traditional gradient operator in the ability to express details. Secondly, a new anti-noise coefficient based on Gaussian curvature and noise intensity is proposed to solve the problem that the meter gradient operator is allergic to large noise points. In addition, a reverse diffusion filter based on local variance of residuals is introduced to enhance the smoothed texture information in the image. Finally, the new model is discretized by finite difference algorithm, and simulation results show that the proposed ANRDPM model not only performs well in smoothing image noise, but also effectively protects image texture information and structural integrity.