Abstract: Social media sharing websites like Flickr allow users to annotate imageswith free tags, which significantly contribute to the development of the webimage retrieval and organization. Tag-based image search is an important methodto find images contributed by social users in such social websites. However,how to make the top ranked result relevant and with diversity is challenging.In this paper, we propose a social re-ranking system for tag-based imageretrieval with the consideration of image’s relevance and diversity. We aim atre-ranking images according to their visual information, semantic informationand social clues. The initial results include images contributed by differentsocial users. Usually each user contributes several images. First we sort theseimages by inter-user re-ranking. Users that have higher contribution to thegiven query rank higher. Then we sequentially implement intra-user re-rankingon the ranked user’s image set, and only the most relevant image from eachuser’s image set is selected. These selected images compose the final retrievedresults. We build an inverted index structure for the social image dataset toaccelerate the searching process.
Keywords: Re-ranking,Social Clues, Tag-based Image Retrieval, Social Media, Image search.