Abstract:

This paper mainly focuses on application for non-rigid contour tracking in heavily cluttered background scenes. Based on the properties of non-rigid contour movements, a cascading framework for estimating contour motion and deformation is proposed. We solve the non-rigid contour tracking problem by decomposing it into three sub problems: motion estimation, deformation estimation, and shape regulations. First, we employ a particle filter to estimate the global motion parameters of the affine transform between successive frames. Then we generate a deformation probabilistic map to deform the contour. To improve the robustness, multiple cues are used for deformation probability estimation. Finally, we use a shape prior model to constrain the deformed contour. This enables us to retrieve the occluded parts of the contours and accurately track them while allowing shape changes specific to the given object types. Our experiments show that the proposed algorithm significantly improves the tracker performance.