Layout-expectation-based Model for Image Search Re-ranking

Abstract

In this paper, a new algorithm is proposed for image re-ranking in Web search applications. The proposed algorithm introduces a new layout expectation model for improving the image search results. The motivation for using the expectation model is that users may often have potential expectations about the desired image during the search process. By including the layout expectation model to describe users' expectation on image layouts, the re-ranked search results can become more satisfactory to users. Experimental results demonstrate that our proposed algorithm can significantly improve the re-ranking precision compared with the state-of-the-art algorithms.