High probability impulsive noise removal for corrupted images

We know one characteristic of image is that each pixel is very close to surrounding neighbors. If the impulsive noise appears in the local area, it will be very abruptly. In the proposed scheme, we pick up several neighboring pixels around the center pixel in the scanning window as the statistic factors. The mean value and mean of absolute neighboring difference (MAND) are utilized from those. Further, we use them to acquire a critical value. The value is a representative characteristic of those pixels in the local area and it is used to identify whether if a pixel is corrupted.