Time Based Re-ranking for Web Image Search

Search re-ranking is considered as a best way to improve retrieval precision. This paper addresses two common problems in search. The top-ranked results for such queries may not contain relevant image to the user’s search result, and fresh and relevant pages may not get high ranks because when we retrieve and rank, images occur mostly based on the visit count of the link. We introduce a new ranking method named Meta rankers, which is based on Time, Download and also View. We invoked ranking based on time count of each visit of the image. Image redundancy is still a problem area concerned. We proposed Image check algorithm, which is used to eliminate the redundancy of an image. In this image check algorithm, it compares the color value of every single expected and actual pixel and avoids the repetition of the same image. Therefore, it provide better and most engaging and diverse search results.

Download PDF