Research on a Distributed Approach for Large-scale Image Retrieval based on Location-sensitive Hashing

Abstract

With the rapid development of network and multimedia technologies, a large number of image databases have been produced, and image data has been experiencing an exponential growth. Meanwhile, the requirement for searching images, i.e. image retrieval in large databases is a hot issue. Many traditional methods search for images accurately according to the feature vector of the image, but this method is computationally expensive. The most famous improved solution is the Locality Sensitive Hashing (LSH, Locality Sensitive Hashing) method which sacrifices some search accuracy to improve efficiency. However, most of the current LSH methods are difficult to handle large-scale data. Therefore, in this paper we design and implement a distributed image retrieval approach based on LSH. The hash value of an image is used to filter out a batch of completely irrelevant pictures, and then the distributed computing resources are utilized to find similar pictures. Experimental results show that the accuracy rate and recall rate could reach about 90%, at the same time the retrieval speed is acceptable, about 120 ms for each retrieval on average.