ABSTRACT
Video analysis problems have been the subject of extensive research in the past [37] [34] [26]. One of these problems is video background modeling and foreground detection, which is actually the process of separating out foreground objects from the background in a sequence of video frames. It is a crucial task in computer vision and has been employed in many video applications involving video surveillance, traffic monitoring, human motion analysis and object tracking. Video background subtraction is a significant challenging problem due to the dynamic nature of video backgrounds generally characterized by changing illumination levels, temporal background clutter as often found in outdoor scenes and non-stationary background objects such as rain, snow, moving leaves and shadows cast by moving objects. In recent years, many research efforts have been devoted to the study of video background subtraction and different techniques have been deployed [2] such as high level region analysis [36], kernel density estimation [12], Markov random fields [27], and hidden Markov models [33]. All of the aforementioned approaches have their own advantages and shortcomings. A review of background subtraction techniques can be found in [28].
