As a rising technology to help adaptable content based image reterival (CBIR), hashing has as of late gotten extraordinary consideration and turned into an extremely dynamic research area. In this investigation, we propose a novel unsupervised visual hashing approach called semantic-assisted visual hashing (SAVH). Recognized from semi-administered and directed visual hashing, its center thought is to viably remove the rich semantics idly implanted in helper writings of pictures to support the adequacy of visual hashing with no express semantic names. To accomplish the objective, a bound together unsupervised system is produced to learn hash codes by all the while saving visual likenesses of pictures, coordinating the semantic help from helper messages on demonstrating high-arrange connections of between pictures, and portraying the relationships among’s pictures and shared themes. Our execution think about on three openly accessible image accumulations: Wiki, MIR Flickr, and NUS-WIDE shows that SAVH can accomplish better execution more than a few best in class procedures.