Parts-based shape recognition via shape geodesics

The quality of the segmentation process directly affects the performance of the shape recognition. In this paper, we address the problem of shape recognition using only the available shape parts instead of the whole shape. For this purpose, we propose a shape parts recognition strategy that uses a robust distance based on geodesics in the shape space. The proposed combining strategy seeks to handle the contour discontinuity can occur in edge maps due to various disturbing factors encountered in real images. The experimental validation through the MPEG-7 shape database and some real images demonstrates the efficiency of our proposed approach.