Background subtraction using running Gaussian average and frame difference

Abstract

Background Subtraction methods are wildly used to detect moving object from static cameras. It has many applications such as traffic monitoring, human motion capture and recognition, and video surveillance. It is hard to propose a background model which works well under all different situations. Actually, there is no need to propose a pervasive model; it is a good model as long as it works well under a special situation. In this paper, a new method combining Gaussian Average and Frame Difference is proposed. Shadow suppression is not specifically dealt with, because it is considered to be part of the background, and can be subtracted by using an appropriate threshold. At last, a new method is raised to fill small gaps that the detected foreground or the moving objects may contain. © IFIP International Federation for Information Processing 2007.