A soft weighted median filter for removing general purpose image noise

In this paper, we propose a new simple method for removing all types of image noise using Soft Weighted Median Filter (SWMF). For each pixel in image, firstly we determine the window to search neighboring pixels. Then all the pixel values in the window are sorted and divided into three parts, the pixel value in part two are considered noise-free pixels, while the pixel values in part one and part three are considered noise candidates. Finally, the noise candidate is replaced by the weighted median filter value of the current window, while the noise-free pixels are left unchanged. The experimental results on ten images with different characteristics indicate that our proposed method is successful for all noise types with an average 24.26 db PSNR value and high restoration quality especially for gaussian, poisson, speckle, and localvar noise.