A New Background Subtraction Method Using Bivariate Poisson Process
Thi Thi Zin, Pyke Tin, Takashi Toriu, Hiromitsu Hama · 2014
Background subtraction is one of important fundamental steps in many image processing applications such as object recognition, detection, tracking, human behavior analysis in video surveillance systems, etc. So the background subtraction method must be efficiency, that is saving time and space and have a good performance. In order to achieve this aim, a new background subtraction method is proposed by using a bivariate Poisson process which takes both serial and spatial correlations of image pixels. The proposed method can deal with complex background scenarios including slowly moving foreground objects, illumination changes and etc. Numerous experiments on various types of video sequences show that the method is robust to compare with several existing methods, can achieve very promising performance.