Symmetrical recursive median filter for region smoothing without edge distortion

In this paper, we present a new median-based operator which associates in a multidirectional filtering scheme the noise suppression ability of the recursive median filter with the edge conservation property of the standard median filter. We show in a preliminary comparative study the superiority of the recursive median filter for smoothing regions corrupted by additive white noise, while the standard median filter is showed to be better in preserving the edges with less distortion. A new operator called the symmetrical recursive median filter is then introduced aiming to improve noise suppression properties as well as to conserve accurately object contours in the filtered images. The properties of the proposed operator are discussed. The operator effectiveness for region smoothing without edge distortion is illustrated on real and synthetic image processing examples.