Abstract
Video reranking is an effective way for improving the retrieval performance of keyword-based video search engines. A fundamental issue underlying the success of existing video reranking approaches is the ability in identifying potentially useful recurrent patterns from the initial search results. These patterns can be leveraged to upgrade the ranks of visually similar videos, which are also likely to be relevant. However, mining useful patterns without understanding query may lead to incorrect judgment in reranking. We explore the user selected data, which can be viewed as the footprints of user searching behavior, as an effective means of understanding query, for providing the basis on identifying the recurrent patterns that are potentially helpful for reranking. In this paper, a new reranking algorithm, named user feedback assisted multi-modality reranking, is proposed. The algorithm leverages selected videos to locate similar videos that are not selected, and reranks them in a multi-modality learning scheme. Experimental results obtained by applying the proposed method to a real-world video collection show its effectiveness.
