Structural Adaptation for Patch-Based Image Denoising

Abstract

Most existing patch-based denoising methods are based on fixed size patch and fixed range of neighborhood, while producing state-of-the-art denoising results, the quality of the algorithm is significantly dependent on patch size. Suppose that patch size is related to image pattern, a structural adaption method for the selection of patch size is proposed in this paper. Due to the mutual similaritybetween the training samples, the simulation results illustrate the original approach can be effectively improved by adopted this method.