A Modified Directional Weighted Median Filter using second order difference based detection for impulse noise removal

This paper proposes a method for impulse noise removal using a Modified Directional Weighted Median Filter (MDWM). The proposed approach has two phases. In the first phase, corrupted pixels are identified using second order difference based detector. In the second phase, MDWM is applied to remove noise. MDWM is an improved directional weighted median filter, which replaces only the corrupted pixels in the image, leaving uncorrupted pixels unchanged. The corrupted pixel is replaced by the median of the pixel values in all the four main directions in a window. These pixels are associated with a weight value for median calculation. Here, maximum weights are assigned to the pixels in the direction with minimum deviation. The proposed method had been tested on benchmark images. Experimental results show the superiority of the proposed method in terms of PSNR, IEF, SSIM and IQI for a few iterations.