Relevancy tag ranking

Abstract

Tags are metadata which helps describe the visual content of an image that makes browsing easier by better organization. Recent boom of Social Media sharing Websites has popularized tagging among a large pool of users by facilitating sharing and embedding of personal photographs. Inappropriate and Random tagging has come out of the blue as a major drawback of personalized tagging limiting the effectiveness of their search and retrieval. In this paper, we propose a tag indexing scheme, which helps to rank the tags of an image according to their pertinence with image content. We first segment the image, calculate the size of segmented objects, and continue parsing for object identification. Then we perform the Probabilistic density estimation and finally couple it with social image retrieval approaches to improve its effectiveness. This tag ranking approach significantly hikes up the performance of tag based image search and retrieval.