Improved adaptive spatial distance-weighted median filter

Abstract

This paper introduces a new class of weighted median filters. The proposed algorithm employs a spatial distance weighting function, which is based on the conclusion that the degree of similarity between two stimuli can be quantified as a simple exponential decay function of a normalized distance in a psychological space. Depending on two parameters of the weighting function, the proposed approach can provide filtering performance ranging from an identity operation to that of the standard median filter (SMF), and by adaptively adjusting one of the parameters, the best possible filtering effect may be achieved. The experimental results show the supriority of the proposed solution to the SMF and some other median filtering methods in terms of both the objective measures and the perceptual visual quality.