A decision based switching median filter for restoration of images corrupted by high density impulse noise

Digital images are corrupted by impulse noise mainly due to sensor faults of image acquisition devices and adverse channel environment which in turn degrades the image quality. A decision based switching median filter (DBSMF) to restore images corrupted with high density impulse noise is proposed in this paper. The global use of standard median filters for impulse noise removal from corrupted images provide good results but the filtering operation may affect fine pixels in addition to noisy pixels which leaves a blurred effect on the filtered image. In order to address this issue the proposed algorithm makes use of an efficient detection scheme to identify the noise pixels and noise free pixels. The detection algorithm clusters the pixels in the corrupted image so as to fall under three categories which states whether the pixels are corrupted or uncorrupted. The proposed switching median filter processes only on those pixels that are classified as corrupted and replaces the processing pixel by the median value. Under high noise densities the filtering window consists of more number of corrupted pixels. For such cases, the proposed algorithm restricts certain conditions on the expansion of the filtering window size to effectively choose the median value. The performance of this decision based algorithm is tested against four noise models for different levels of noise densities and is evaluated in terms of performance metrics which include Peak Signal to Noise ratio (PSNR) and Image Enhancement Factor (IEF). It gives better results for images that are extremely corrupted up to 90% noise density and outperforms classic filters in terms of handling image corruption.