Joint Coupled-Hashing Representation for Cross-Modal Retrieval
Yihan Liu, Zhaojia Chen, Cheng Dan Deng, Xinbo Gao · 2016
Cross-modal retrieval based on hashing has attracted much attention for its low storage cost and fast query speed. Most cross-modal hashing methods arbitrarily embed the heterogeneous features into a common Hamming space, which is unsuitable for many real-world applications and makes the search results unsatisfactory. In this paper, we propose a novel cross-modal hashing method, which embeds heterogeneous features into their respective Hamming spaces. The proposed method has two advantages: 1) reconstructive embedding in each individual modality represented by matrix decomposition is performed to enhance the discriminative ability of the binary codes; 2) the correlations between different modalities are maximized by joint coupled-hashing representation. Experimental results on two public datasets demonstrate that our method gains better retrieval performance than several state-of-the-art approaches.