The problem of computational model and algorithm for extracting and tracking objects moving in complicated scenes has been studied from the viewpoint of visual perception. The effect of changing model parameters on optimum threshold selection for image segmentation has been analysed and tested.It is different to conventional algorithms that the new approach proposed integrates proper object background condition
visual nonlinearity
interframe correlativity and differences into a two-step segmentation process which includes three criteria and a fast optimizing procedure.Object tracking is then implemented by binary template matching.The experiments conducted on visibleband image sequences are reported.