An Improved Reaction-Diffusion Filters for Denoising

摘要

Motivated by the classical TV (total variation) restoration model, we propose a new nonlinear reaction-diffusion digital filter based on Nordstrom energy functional for denoising and enhancing digital images, or more generally, data living on graphs. Using a new weight function, the filter is a data dependent low-pass filter, capable of denoising data without blurring jumps or edges. Numerical results show that the model is not only for images with Gaussian noise, Salt & pepper noise or Speckle noise, but also suitable for mixed noise, the denoised image has higher PSNR and good visual effect.