Description

A content-based image retrieval (CBIR), hashing has been recently received great attention and became a ver y active research domain. In this study, we propose a novel unsuper vised visual hashing approach called semantic-assisted visual hashing (SAVH). Distinguished from semi-super vised and super vised visual hashing, its core idea is to effectively extract the r ich semantics latently embedded in auxiliar y texts of images to boost the effectiveness of visual hashing without any explicit semantic labels. To achieve the target, a unified unsuper vised framewor k is developed to lear n hash codes by simultaneously preser ving visual similar ities of images, integrating the semantic assistance from auxiliar y texts on modeling high-order relationships of inter-images and character izing the correlations between images and shared topics.