A novel approach to remove random-valued impulse noise from digital image

In the field of switching filter, a highly effective filter to restore extremely corrupted image with impulse noise is presented. It is capable of handling low density as well as high density of random valued and fixed valued impulse noise. In this study, local area comprises within the window in an image is analyzed for intensity extrema to classify the pixel as either noisy or noiseless. Filtering is applied to the noisy pixels only and it is done in such a way that the noisy pixel is replaced by either the median or the mean value of the filtering window depending on the noiseless pixels present in the window. The window size is adaptive for this filter and depends on the estimated noise density. The proposed filter is tested on a large number of grayscale and color images under a wide range of noise density (from 10% to 94%) and the simulation results reveal that it performs better than other approaches to impulse noise removal, in terms of suppressing impulse noise while preserving image details. The proposed filter is simple to implement and suitable for real time implementation.