Reduction of Salt-and-Pepper Noise from Digital Grayscale Image by Using Recursive Switching Adaptive Median Filter

Abstract

Digital images may suffer from impulse noise, including salt-and-pepper noise. One of the common methods to deal with this noise is by using median filter, which is a type of non-linear filter. Standard median filter includes noisy pixels in calculating the median value for the restoration process. However, this will lead to an inaccurate result, where the noisy pixel values may be selected for the restoration. Another approach is by using recursive median filter, where the calculation for the median value is also based on the previous outputs. Therefore, in this paper, we investigate the feasibility of improving the performance of recursive median filter, by adapting it to switching and adaptive approaches. This scheme is called as Recursive Switching Adaptive Median Filter. As the switching median filter is used, the method is divided into two stages, which are noise detection and noise restoration stages. In the noise detection stage, salt-and-pepper pixel candidates are identified. Then, in the restoration stage, an adaptive method is used for the restoration. The size of the filter is expanding until there are at least eight noise-free pixel candidates defined by the window. As the recursive method is used, the noise mask is updated every time the restoration is done. The experimental results show that this scheme has good performance in terms of mean square error and structural similarity index measure, as compared to six other median filtering approaches. However, the scheme does not perform well at high level of corruption, especially when the level of corruption is more than 80%.