Topic hypergraph hashing for mobile image retrieval

Hashing is one of the promising techniques to support efficient Mobile Image Retrieval (MIR). However, most of existing hash-ing strategies simply rely on low-level features, which inevitably makes the generated hashing codes less discriminative. Moreover, many of them fail to exploit complex and high-order semantic cor-relations between images. Motivated by these observations, we propose a novel unsupervised hashing scheme, called Topic Hyper-graph Hashing (THH), to address these limitations. A unified topic hypergraph, where images and topics are represented with inde-pendent vertices and hyperedges respectively, is first constructed to model latent semantics of images and their correlations. With topic hypergraph model, hashing codes and functions are then learned by simultaneously preserving similarity consistence and semantic cor-relation. Experiments on standard datasets demonstrate that THH can achieve superior performance compared with several state-of-the-Art techniques, and it is more suitable for MIR.