Clustering based one-to-one hypergraph matching with a large number of feature points

Hypergraph matching is a useful technique for multiple feature point matching. In the last decade, hypergraph matching has shown great potential for solving many challenging problems of computer vision. The matching of a large number of feature points in hypergraph constraints is an NP-hard problem. It requires high computational complexity in many algorithms such as spectral graph matching, tensor graph matching and reweighted random walk matching. In this paper, we propose a computationally efficient clustering based algorithm for one-to-one hypergraph matching, which clusters a large hypergraph into many sub-hypergraphs. These sub-hypergraphs can be matched based on a tensor model, which guarantees the maximum matching score. The results from the sub-hypergraphs are then used to match all feature points in the entire hypergraph. Simulation results on real and synthetic data sets validates the efficiency of the proposed method.