Hashing has drawn increasing attention in cross-modal retrieval due to its high computation efficiency and low storage cost. However, there is a certain lack in the previous cross-modal hashing methods that they can not effectively represent the correlations between paired multi-modal instances. In this paper, we propose a novel Hypergraph-based Discrete Hashing (BGDH) to solve the limitation. We formulate a unified unsupervised hashing framework which simultaneously performs hypergraph learning and hash codes learning. Hypergraph learning can effectively preserve the intra-media similarity consistency. Furthermore, we propose an efficient discrete hash optimization method to directly learn the hash codes without quantization information loss. Extensive experiments on three benchmark datasets demonstrate the superior performance of the proposed approach, compared with state-of-the-art cross-modal hashing techniques.
展开▼
