Object Motion in Bayesian Propagation of Ovals

Many tracking applications seek essentially the whereabouts of the object of interest, its rough location and shape size rather than its precise body outline. This often relieves the problem of much of the computing complexity. We here propose a tracking method that is based on approximating with an oval the moving object in a video sequence of moving background. Through the use of the proximate distribution densities of the local regions, the discriminating features of the object are extracted from a small neighborhood of the local region containing the tracked object. By estimating the object's location probability in a Bayesian framework, we identify the object via an approximating oval, thus using the ovals to trace the object motion. The method remains effective even when there are certain object occlusion, and illumination and shape changes.