An Efficient Directional Weighted Median Switching Filter for Impulse Noise Removal in Medical Images

Madhu S. Nair, J. Reji

Abstract: This paper proposes a new efficient directional weighted median filter for impulse noise removal in medical images. The proposed method consists of two phases: noise detection and noise filtering. In this method, detection is done by Directional Weighted Median (DWM) detection [1], and filtering is applied to only corrupted pixels in the noisy image. The noise detection stage is based on the differences between the current pixel and its neighbours aligned in the main four directions in the considered window. The weighted median filter is then applied on directional pixel values as well as on uncorrupted pixels in the selected window by giving appropriate weights for the better restoration of corrupted medical images. Extensive experimental analysis shows that the proposed technique can be used for medical images with different impulse type noises. Both quantitative and qualitative analysis shows the superiority of the proposed method over other filters.