A Novel Shape Matching Method Based on Feature Points

Shape matching is an important ingredient in object recognition, classification and image retrieval. Many features have been designed to improve the performance of similarity measure between pairs of shapes. This paper proposes a feature point based method using shape context descriptor. To improve the accuracy of shape matching, the proposed method incorporates appearance similarity constraint into the correspondence estimation. The optimal point correspondences between two shapes then can be obtained by minimizing the cost of matching points between two shapes. On the other hand, to reduce the computational cost on shape matching, we utilize feature points to locate the corresponding sampling points. The proposed algorithm is evaluated on two famous databases, i.e., the database of MPEG-7 shape database and the coil-100 image database. The experimental results demonstrate that the proposed algorithm is robust and efficient in shape matching.