Video Clip Retrieval Based on LBP Variance

Abstract Advancement in technology has made the acquisition and storage of multimedia data easy and inexpensive to the end user. However for effective use of the information available in the multimedia, efficient and accurate retrieval methods are required. Multimedia based retrieval systems extensively used texture based approach to interpret and recognize a scene image. The texture of an image provides clue to the orientation, smoothness, symmetry, shape, regularity and coarness of the surface. The Local Binary Pattern Variance (LBPV) is a texture feature where variance in contrast acts as adaptive factor during computation of local binary pattern (LBP) histogram. The LBP combines both structural and statistical approaches to texture analysis. This paper proposes an LBP Variance based approach to visual content based video retrieval. The proposed approach uses query by example paradigm for retrieving similar clips from the video. Experiments conducted on TRECVID dataset shows the efficacy of proposed approach.