Human body detection and tracking in video sequences using chamfer matching

We describe a chamfer based matching algorithm which efficiently search the best match from high number of human templates using distance transformation. The novel part of the algorithm is that we introduce a template space which offers more information to create an optimal templates hierarchy and it can be used directly to track people more efficient as the coarse-to-fine technique. This matching method is also eligible to give the pose of the detected people. In this paper we also present an efficient pseudo parallel method to compute the distance transformation.