Adaptive window multistage median filter for image salt-and-pepper denoising

The median filter, as the most common one of nonlinear filters, possesses the incomparable advantage over linear filters - overcome the edge blur. Many improved versions have been developed to reduce noises and reserve details. They are aimed at the images containing much detail or corrupted by high percentage of noises. However, none of currently proposed median filters could provide perfect adaptability to both sides. The proposed method is aimed at the former case. By analyzing eight kinds of noises and signals, we proposed an algorithm called adaptive window multistage median filter for image salt-and-pepper denoising. Experiments results demonstrated its superiority over other methods when the percentage of salt-and-pepper noise is less than 20.