Srdjan Stanković, University of Montenegro, Faculty of Electrical Engineering

Compressive sensing inspired image denoising

Abstract

This presentation reviews image denoising approaches that explore sparsity property and reconstruction principles used in compressive sensing conditions. In the light of compressive sensing, image denoising is observed as a process of recovering the pixels corrupted by noise while assuming that these pixels are unavailable. In other words, compressive sensing can be efficiently employed as a tool for image denoising. Here we observe two common noise types, namely, the impulse noise (salt and pepper) and the Gaussian noise. When dealing with impulse noise, the robust estimators are found to be productive in estimating the positions of corrupted pixels that are further processed as the missing ones. The challenge seems to be greater in the case of Gaussian noise, where denoising is performed iteratively. In each iteration, a certain set of pixels is declared missing and then reconstructed. It has been shown that such approach can reduce the noise through the iterations rather than to remove it completely. As a more general case, we present the method for detecting and recovering the noisy pixels when the noise is mixed and some of the noise amplitudes are within the range of the non-noisy pixels values. The noisy pixels are detected using the gradient of a sparsity measure as a criterion. Namely, the noisy pixels mainly degrade the image sparsity in the transformation domain which is used to select and discard all corrupted image pixels. The discarded pixels are recovered using the gradient algorithm for compressive sensing reconstruction. Herein the discarded noisy pixels are considered as the minimization variables, while the minimization problem is formulated using the gradient of the L1-norm.

Short Biography

Srdjan Stanković was born 1964 in Montenegro. He received the B.S. (Hons.) degree in Electrical Engineering from the University of Montenegro, in 1988, the M.S. degree in Electrical Engineering from the University of Zagreb, Croatia, in 1991, and the Ph.D. degree in Electrical Engineering from the University of Montenegro in 1993. He received the Awards for the first place in Montenegro in the High School Physics competitions (1981, 1982), the Award for the best student in Montenegro (1987), the Award of Montenegrin Academy of Science and Arts for scientific achievements as young researcher (1995). From 1988 to 1992, he worked in the Aluminum Plant of Podgorica as a Research Assistant. In 1992 he joined the Faculty of Electrical Engineering, University of Montenegro, where he is currently a Full Professor. In the period 2007-2013 he was the Dean of the Faculty of Electrical Engineering, University of Montenegro. His interests are in Signal processing, Multimedia systems, and Digital electronics. During the winter semester 2014/2015 Prof. Stankovic was on the position of Vice Rector of the University of Montenegro.
He is a member of the Board of Directors in Montenegrin Broadcasting Company since 2004. He was the President of the Board of Directors in Montenegrin Broadcasting Company (2005-2006).
In 1998 he spent a period of time with the Department of Informatics at the Aristotle University in Thessaloniki, supported by Greek IKY foundation. In the 1999-2000, he was on leave at the Darmstadt University of Technology, with the Signal Theory Group, supported by the Alexander von Humboldt Foundation. In 2002, he spent three months at the Department of Computer Science, the University of Applied Sciences Bonn-Rhein-Sieg, as an Alexander von Humboldt Fellow. From 2004 to 2006, he stayed several times with the E3I2 Laboratory, ENSIETA, Brest, France. From 2007 to 2010 he visited (one month research stay) Centre for digital signal processing research at King's College London, Laboratory of mathematical methods of image processing, at Moscow State University Lomonosov, as well as GIPSA Laboratory at INP Grenoble. He spent academic 2012/2013 with the Center for Advanced Communications at the Villanova University, PA. In 2014 he spent a semester with GIPSA Lab, INP Grenoble. In 2016 he stayed (one month stays) at the Swanson School of Engineering, University of Pittsburgh as well as at the Heidelberg Colaboratory for Image Processing, Heidelberg University He has published a book "Multimedia Signals and Systems" by Springer and several textbooks on electronics devices (in Montenegrin). He coauthored three monographs on time-frequency signal analysis (in English). He published more than 250 papers, among them about 90 in the leading scientific journals in the areas of Signals, Systems and Image processing. In 2010, he was the Lead Guest Editor of the EURASIP Journal on Advances in Signal Processing for special issue: Time-frequency analysis and its applications to multimedia signals, as well as the Guest Editor of the Signal Processing for special issue: Fourier related transforms. He was the Lead Guest Editor of the IET Signal Processing for the 2014 Special issue: Compressive Sensing and Robust Transforms. From 2005 to 2009 Dr. Stanković was serving as an Associate Editor of the IEEE Transactions on Image Processing. He is a Senior member of the IEEE. In 2011 he was awarded by the Ministry of Science in Montenegro as the Leader of the Best Scientific Project in Montenegro.