Progressive decision-based mean type filter for image noise suppression

Abstract

A novel progressive decision-based mean (PDM) filter is proposed to restore images corrupted by random-valued impulse noise. An impulse detection algorithm based on the Dempster–Shafer (D–S) evidence theory is used before filtering. This work presents a new approach to automatically determine mass functions for the D–S evidence theory using the feature information provided by the filter window. Decision rules can determine whether noise exists based on the noise-corrupted belief value. The impulse detection and the noise filtering procedures are progressively applied through several iterations. Finally, the input pixels are identified as either noise-free or noise-corrupted, and only the noise-corrupted pixels in corrupted images are replaced by the mean value of the noise-free pixels in the filter window. Extensive simulation results have demonstrated that the proposed algorithm significantly outperforms other median-based filters.