Coupled dictionary learning with common label alignment for cross-modal retrieval

Abstract

Cross-modal retrieval has been an active research topic in recent years. However, most existing methods ignored discovering the common semantic relationship among different modalities so as to seriously reduce the retrieval accuracy. To cope with this problem, we propose a novel cross-modal retrieval method based on coupled dictionary learning with common label alignment. Concretely, our method first conducts coupled dictionary learning on the data from different modalities separately and then projects them into a common space, where the correlation between these modalities is encouraged by using common label alignment. Experimental results on two public datasets demonstrate that our method outperforms several state-of-the-art methods.