A new patch-based approach for noise filetring

Abstract

A common framework is proposed for Gaussian, unilateral, bilateral and non-local means filters. A new denoising method is then introduced for removing additive independent Gaussian noise. It is adapted and applied for filtering signal-dependent noise such as film grain noise. The main idea is to transform the signal-dependent into independent additive Gaussian noise using a nonlinear transformation prior to filtering. The obtained results clearly demonstrate the efficiency of the proposed method.