An Axis Based Mean Filter for Removing High-Intensity Salt and Pepper Noise

Abstract

In this work, we propose an Axis Based Mean Filtering (ABMF) method for removing high-intensity Salt and Pepper Noise from gray-scale images. The proposed method applies the concept of mean filter and uses only the terminal pixels within a window along a specific axis to predict the value of the central noisy pixel. The ABMF considers a fixed window size of 3×3. If the central pixel of the window is noisy, it tries to identify a straight line (referred to as axis) within the window passing through the central pixel such that the pixels on either side of the line are non-noisy. If such an axis is found, the noisy pixel is replaced by the mean of the pixels on either end of the axis. However, if such a line does not exist, the noisy pixel is replaced by the mean of all the non-noisy pixels within the window. Experimental results over a set of 34 images exhibit that the proposed ABMF outperforms the existing algorithms by 58% and 29% in terms of mean SSIM and mean PSNR respectively for noise-intensities ranging from 10%-90%.