Salt and pepper noise removal for image signals

In this paper we propose a new method to recover images corrupted with salt and pepper noise. The proposed method uses ordered statistics filtering and soft-decisioning in successive approximations to find the locations and amplitudes of all the impulses. In our method we use soft-decision to generate a mask which is based on histogram distributions in local windows to distinguish impulsive noise from legitimate pixels. This mask is used to obtain an estimate of the original image by an iterative method. The result of the aforementioned algorithm can be fed back into the system to improve the image reconstruction process. This algorithm is analyzed and verified by simulation results. Computer simulations verify substantial improvement over other methods in terms of peak-signal-to-noise-ratio (PSNR) and visual quality.