A decision-based filter for removing salt-and-pepper noise

Although the traditional median filter and its available variants are usually able to remove salt-and-pepper-noise, filter performances in terms of noise removal and retention of image information decrease as the amount of contaminating noise increases. This paper proposes a simple and effective method to remove salt-and-pepper noises from different types of images. The proposed algorithm replaces a noisy pixel value by the median value in its neighborhood only if the median comes from a noise-free pixel. Otherwise, it takes some special cares so that a noisy pixel value is not replaced by a value coming from another noisy pixel. Finally, it applies robust estimator to smooth the filtered output. The proposed technique is effective even if the image is highly contaminated with noise. The simulation also proofs the superiority of the proposed method above the existing noise-removal techniques.