Image Diffusion In Connection With Robust M-Estimators

Abstract

In this paper, we introduce a nonlinear diffusion method for image denoising using robust M-estimators. In the proposed diffusion model, the diffusivity function is replaced by robust M-estimators weight function and the modulus of gradient in a diffusivity function is substituted by the average of local gradient in a 3x3 window. The proposed method is tested on a number of benchmarks and compared to Weickert‘s diffusivity function. The method reports on high visual quality and high PSNR measures compared to Weickert‘s diffusivity.