An improvement of an adaptive weighted mean filter using fuzzy clustering

Abstract

This paper proposes a novel edge-preserving adaptive weighted mean filter using fuzzy clustering. An input vector in the filter mask is classified according to predefined clusters and the membership values corresponding to all clusters are obtained. The filter output is given by the weighted sum of the membership values with the inner products of the input vector with weight vectors according to the clusters. The proposed filter can reduce mixed noises with preserving edges satisfactory, because a fuzzy clustering flexibly classifies ambiguous local image information and adaptively controls filter weights.