Real-time shot transition detection in compressed MPEG video streams

Abstract

Digital libraries became necessary vehicles that provide users with powerful and easy-to-use tools for searching, browsing, and retrieving media information. The starting point for these endeavors is the same: segmentation of video material into shots. The aim of this study is to segment MPEG video streams into shots. A fully automatic detection for both abrupt and gradual transitions (dissolve and fade groups) with minimal decoding in real time is developed. Each detection process was explored through two phases: macroblock type analysis in bidirectional predictive pictures (Bframes), and on-demand intensity information analysis. The abrupt transition detection is explored first by examining the number of forward and backward macroblocks (p- and b-MBs) in consecutive B-frames, and then an intensity histogram comparison is applied to confirm detected transitions. The gradual transition is detected first by examining the intracoded predicted macroblocks (i-MBs) within successive B-frames, and then the detection is confirmed by checking the parabolic shape of the frame variances of the candidate sequence. Results of the study show remarkable detection rate for both abrupt and gradual transitions.