Publication

Abstract : Background subtraction being one of the most crucial steps in numerous real-world video-based applications has been studied extensively to date. Several researchers have provided solutions to this problem by exploiting the patterns and information hidden in a set of initial raw videoframes with the help of sophisticated machine learning techniques. While others have proposed to uncover the hidden patterns in raw data by employing a variety of texture patterns, which now have become an inseparable part of the background subtraction problem. Hence, the paper proposes to employ a novel feature combination of Attractive–repulsive local binary gradient contours with RGB channels and luminance information into a consensus-based modeling technique. The results obtained on sample videos from the standard Change Detection dataset [1] support for the superiority of the proposed methodology.