Hyper-graph matching algorithm describes the whole structure of object by high-order topology. Previous work has presented many methods to build and solve the problem model. This paper mainly focuses on feature description and result optimization. First, we combine several features in stable region of object as bundled feature, it can describe more relationship by hyper-graph model. Second, we properly extend previous work to build and solve the hyper-graph model. Finally, we optimize the matching result by iteration and modifying constraints, it improves the accuracy effectively. Comparative experiments verify the good performance of our algorithm, especially for non-rigid object matching.
展开▼
