Background modeling in video surveillance by using parallel computing

Abstract

In  the last years , the research of extraction the movable object from video sequence in  application  of computer vision become wide spread and well-known . in this paper the extraction  of background model by using parallel computing is done by  two steps : the first one using non-linear buffer to extraction frame from  video sequence depending on  the number of frame whether it is even  or odd . the goal of this step is obtaining initial  background when over half of training sequence contains foreground object . in the second step , The execution time of the traditional K-mean has been improved to obtain initial background through perform  the k-mean by using parallel computing where the time has been minimized to 50% of the conventional time of k-mean.