Restoration method of images corrupted by high probability impulse noise by using fuzzy technique
Mitsuhiro Okano, Akira Taguchi, Mitsuo Sone · 2002
We propose a novel data-dependent median filtering method in order to restore images which are corrupted by 10/spl sim/50% probability impulsive noise. Even if the image which is corrupted by high probability impulsive noise, the degree of impulsive noise is not uniform at each local window. Thus, the suitable window length is changed at each processed point. If the processed point is not corrupted by impulsive noise, the input signal should be not filtered in order to preserve signal details. From these points of view, the proposed filter combines median filters with a lot of different window sizes (1/spl times/1 window is included) for preserving the signal details and removing impulsive noise perfectly.