Low Complexity Motion Detection with Background Modeling

Motion detection is of utmost importance in several applications such as video surveillance, remote sensing and medical diagnosis, and a large number of algorithms have been proposed for implementation. In this paper we focus on low resources scenarios where standard solutions are inapplicable. We present a novel algorithm that exploits spacial correlation of neighbor pixels in an image to get the most accurate view of the moving scene without increasing computational burden. In conjunction with the proposed algorithm we investigate, to the best of our knowledge for the first time, several background models and update policies, comparing available approaches and exploring trade-offs between complexity and detection capability.