A Novel Surveillance Video Processing Using Stochastic Low-Rank And Generalized Low-Rank Approximation Techniques

Abstract

Background extraction and subtraction is a key way for automatic video analysis, especially in the domain of video surveillance. In this paper, we present a non-iterative stochastic technique for background extraction and subtraction from the surveillance video. At the same time, generalized low-rank approximation matrix technique is employed to compress the background subtraction image in transmitting procedure. The simulation results illustrate that the presented technique is effective in separating the background from the activity and more rapid than the method of the latest literature. The employed method is demonstrated that there is low compress ratio which can avoid the restriction of communication bandwidth caused by big data.