Abrupt Shot Change Detection using Midhinge Local Binary Pattern

Video shot change detection is an extremely important task for video analysis and is considered as the preliminary step for effective video summarization, browsing and retrieval. In this work, a new texture based descriptor has been proposed using Midhinge Local Binary Pattern (MHLBP) concept for identification of abrupt transition in videos. The discrimination capability of MHLBP is better than basic LBP and its versions (mean LBP and midrange LBP). MHLBP histogram is constructed by applying midhinge statistics on each mask of video frame. The distance between histogram features of the adjacent frames are evaluated using Euclidean distance. The obtained distance values are subjected to adaptive thresh-old mechanism for identifying of abrupt shot changes in a video. The planned framework is tested on a subset of TRECVID 2001 dataset. The result shows that the proposed method outperforms with other existing shot change detection algorithms in terms of precision, recall and F-measures.