A potential-based approach for shape matching and recognition

This paper presents a novel potential-based approach for recognizing the shape of a two-dimensional (2D) region by identifying the best match from a selected group of shape templates. The proposed model assumes that the border of every 2D region is uniformly charged. An initially small shape template placed inside a shape sample will experience the repulsive force and torque arising from the potential field. A better match in the shape between the template and the sample can be obtained if the template translates and reorients itself to reduce the potential while growing in size. The shape template with the largest final size corresponds to the best match and represents the shape of the given sample. The potential and the associated repulsive force and torque between the polygonal contours are analytically tractable, hence resulting in high computational efficiency of the matching process. The proposed approach is intrinsically invariant under translation, rotation and size changes of the shape sample. Moreover, not only can the matching be carried out directly for shape contours at different viewscales, but the contours can also be unconnected, provided that the template is confined within the shape sample throughout the matching process.