Video shot boundary detection using midrange local binary pattern

The fundamental step in content based video retrieval is shot boundary detection. This step is essential for characterizing videos in any video processing system. Reliable detection of shots in videos is still a challenging issue. In this work, we address automatic detection of abrupt shots in video sequences. We have proposed a method termed as Midrange LBP (MRLBP), which enhances the discriminative capability of basic LBP. Each frame of a video is processed to extract LBP histogram values based on midrange statistics for its description. A dissimilarity measure is applied on the feature vectors of adjacent frames and the distance values obtained are used for shot detection process using adaptive threshold approach. To check the efficacy of the proposed method, we carried out experiments on a subset of standard video data set TRECVID 2001. The results obtained by the proposed approach outperform the existing shot boundary detection algorithms in terms of different performance measures.