Comparison of the PDE-Based Regularization Methods and a Unifying Framework

Abstract

The frequent problems in computer vision consist of de-noising, artifact elimination as well as structure preserving or enhancing. PDE-based nonlinear diffusion filter may be one possibility to achieve those goals. In this paper, we perform comparison of three typical PDE-based regularization algorithms followed by the proposal of a general framework, which exploits fundamental significance for analyzing PDE-based regularization methods.