Background Subtraction in Mobile Cameras by MRF-MAP Based Optical Flows
Limin Zhu, Yue Fei Zhou · 2013
Background subtraction has always been an important field in computer vision since it is generally the first step in video or image processing. It is an even more challenging job when sequences are taken by moving cameras. Unlike static surveillance cameras, the background varies a lot in every frame taken from mobile cameras. In this sense an optical flow based clustering method is proposed. The unique optical flow representing foreground movement is detected. The final output is given after a region growth in a MRF-MAP approach. Our method is evaluated on different datasets to show its effectiveness.