Unrestricted multivariate medians for adaptive filtering of color images

Abstract

Reduction of impulse noise in color images is a fundamental task in the image processing field. A number of approaches have been proposed to solve this problem in literature, and many of them rely on some multivariate median computed on a relevant image window. However, little attention has been paid to the comparative assessment of the distinct medians that can be used for this purpose. In this paper we carry out such a study, and its conclusions lead us to design a new image denoising procedure. Quantitative and qualitative results are shown, which demonstrate the advantages of our method in terms of noise reduction, detail preservation and stability with respect to a selection of well-known proposals.