Planar shape matching and feature extraction using shape profile

In this paper a novel cross correlation technique is proposed for shape matching between two similar objects. The proposed technique can not only evaluate the similarity between any two objects, but also has two distinct advantages compared to previous work: (1) the deformed articulated objects such as human being with different poses, can be matched very well; (2) the local feature extraction and correspondence can be established at the same time. The basic tool we used is the shape profile driven from the curvature map of the object profile. The cross correlation technique is applied to the shape profile of the two objects to evaluate their similarity. Filtering scheme is used to enhance the quality of both shape matching and extracted features. The invariant property, the robustness and the efficiency of the shape profile in shape matching and feature extraction are discussed.