Generic framework for content-based stereo image/video retrieval

Abstract

With the increasing number of stereo images/videos in commercial markets, the demand for content-based image retrieval (CBIR) to deal with stereo media becomes urgent. To meet this demand, a novel framework is proposed where depth cues are extracted from stereo pairs and employed in a re-ranking scheme to refine results from conventional CBIR. Experiments show the proposed method yields promising results in retrieving stereo content. © 2011 The Institution of Engineering and Technology.