Dual window selective median switching filter

Abstract

A simple and efficient nonlinear image filtering technique, called the dual window selective median switching (DWSMS) filter, for the removal of both additive and impulse noise from corrupted images, is proposed. Two moving concentric windows are employed to determine the luminance value for replacing that of the center pixel. Three steps are involved in this filtering technique. First, the usual median filtering is applied to the smaller window. Second, the median filtering is applied to the larger window. Third, the two median filtered values are then compared to the original center pixel, and the median-filtered value closer to the center pixel is chosen to be the pixel's filtering output. Tests to the standard image corrupted with different amounts of additive and impulse noise show that the proposed algorithm consistently provides improved performance to the original median filtering of both window sizes.