Arab J Sci Eng (2015) 40:3233–3246 DOI 10.1007/s13369-015-1799-2 RESEARCH ARTICLE - COMPUTER ENGINEERING AND COMPUTER SCIENCE Adaptive Switching Non-local Filter for the Restoration of Salt and Pepper Impulse-Corrupted Digital Images Justin Varghese 1 · Nasser Tairan 1 · Saudia Subash 2 Received: 1 December 2014 / Accepted: 23 July 2015 / Published online: 12 August 2015 © King Fahd University of Petroleum & Minerals 2015 Abstract The paper presents an effective nonlinear adap- tive switching non-local filter for the restoration of impulse- corrupted digital images by using distinct impulse detec- tion and correction stages. The correction scheme of the fil- ter adaptively switches between details-preserving non-local mode and signal restoration-based local mode to facilitate high fidelity in the restored image. The non-local filtering operation replaces impulses with a remote pixel that better suits the local image conditions. The algorithm works in this non-local mode only when there are sufficient uncorrupted pixels in the local neighborhood of the corrupted pixel to be replaced. Otherwise, the algorithm replaces impulsive pixels with the median of the uncorrupted pixels from the local neighborhood. Experimental results from various impulse noise levels support the improved performance of the pro- posed algorithm over other algorithms both subjectively and objectively. Keywords Impulse noise · Median filter · Adaptive filter · Image restoration · Nonlinear filter 1 Introduction The information in digital images when acquired through digital sensors or while transmitted through physical/ electromagnetic media is impulse-corrupted due to man- made/instrumental errors or environmental interferences and B Justin Varghese justin_var@yahoo.com 1 College of Computer Science, King Khalid University, Abha, Kingdom of Saudi Arabia 2 Centre for Information Technology and Engineering, Manonmaniam Sundaranar University, Tirunelveli, India loses validity. Impulse noise is a rapid rise/fall in pixel inten- sities [1]. Image restoration is important to overcome the adverse impact of impulse noise on these valuable images. The earlier impulse filters in the form of linear mean filtering algorithms [1] blur fine details of restored image with min- imal noise removal. Many transform domain-based filters were developed [2] to effectively restore impulse-corrupted images. Transform-based algorithms include Discrete Fourier Transform [2], Discrete Cosine Transform (DCT) [2], Discrete Wavelet Transform (DWT) [3], Curvelet trans- form [4], anisotropic diffusion [5]-based algorithms. These filters are computationally complex due to the transforma- tion process and are ineffective for impulse noise reduction [6]. Spatial filters dominated the area due to their simplicity and compactness. The nonlinear position-invariant simple median filter (SMF) [1] is the first in the nonlinear spatial domain filter to provide promising results from impulse- corrupted digital images. The extensions of median filter, the minimum and maximum filter [1], rank-ordered mean filter (ROMF) [1], weighted median filter (WMF) [1] and center weighted median filter (CWMF) [2] that evolved later, oper- ated in spatial domain are computationally efficient at all impulsive situations and favored the development of major restoration operators in spatial domain. These nonlinear spa- tial filters though dealt impulsive characteristics, consumed image details since they treated both impulsive and non- impulsive pixels alike and necessitated later algorithms to incorporate switching schemes that apply filtering operation only to the detected noisy pixels [6]. The adaptive median of absolute deviation (MAD)-based threshold filter (AMTF) [7] incorporates an adaptive threshold to MAD-based impulse filtering approach. But at higher noise ratios, the need for the recursive filtering operation increases the complexity of the algorithm. Also, the knowledge of noise-free image for threshold fixing is not available at most of the times. The iter- 123