Query Adaptive Search System Based On Hamming Distance for Image Retrieval

Abstract

The most recent active topic of research for image retrieval is scalable image search based on visual similarity. The main motivation for image retrieval is based on image ranking, given by multiple retrieval methods without affecting their scalability. This paper describes ranking and retrieval as graphs of candidate images and proposes a graph-based query specific rank fusion approach, in which graphs are created by using nearest neighbour node and multiple graphs are merged together and re-ranked them by conducting link analysis on the fused graph. Then rank fusion maintains the efficiency and scalability of image retrieval by applying the rank aggregation method. The proposed system will add the query adaptive image to the search system. In this system, hamming distance is calculated and query adaptive weights are computed between query image and database image. Based on these weights, images are ranked. A finer-grained ranking of search results is produced by query --adaptive approach. The proposed will improve the efficiency and scalability of image retrieval.