Generalized Fuzzy Peer Group for Removal of Mixed Noise from Color Image

Abstract

In this letter, a novel method has been proposed, which includes formulation of similarity function and a fuzzy-based method for filtering mixed noise. The similarity function is adaptive to local noise level and edge information, and it is used to detect similarity among pixels in a peer group. Based on the peer group, color image corrupted with mixed Gaussian and impulse noise is filtered. The novel method for filtering is an adaptive weighted average of different sized filters. The weights of different sized filters are adaptive to local noise and edge information. The proposed work has been compared with some state-of-the-art techniques. The results show proposed approach is better in preserving edge and color information than others.