Self-organizing maps based impulse detector for switching median filters

In this paper, a new impulse detector is proposed in order to realize a switching median filter which can remove salt-pepper impulse noise while preserving details. The detector judges whether a central pixel of a filter window is an impulse pixel or not. In the detector, a pixel pattern on the window is classified into a cluster by using self-organizing maps, and then the difference between the central pixel value and a median of the window is calculated. Finally, detection is achieved by comparing the difference with a threshold value which is assigned to the cluster. Simulation results demonstrate that the proposed algorithm is better than the conventional algorithm and is particularly effective for the cases where the images are very highly corrupted