A Novel Approach for Motion Segmentation in Moving Pictures: Centre of Mass Model

A novel background model for segmentation of moving objects in running images is presented in this paper. The segmentation algorithm is based on the statistics of the current and previous locations of center of mass of super pixels that are the collection of a group of pixels. The algorithm is superior to those previously reported in the literature due to its immunity to sudden illumination changes and slight changes in the intensity distribution functions of the super-pixels. The performance of the center of mass model is compared with previously reported seven different models that were selected from the background subtracting methods and the statistical methods.