SHOT BOUNDARY DETECTION USING HILBERT TRANSFORM BASED FEATURES

The first and foremost step in the shot boundary detection is the extraction of features from the video sequences. To obtain better performance in shot boundary detection, a new method is proposed in this paper, where texture and local binary information extracted from the hilbert transformed frames are processed and represented as features. The similarity between the frames is constructed as continuity signal. The boundaries between the shots are identified by applying the shot transition identification procedure on the continuity signals. The proposed method is evaluated over the TRECVID 2007 SBD dataset and the performance is compared to the top performers of TRECVID 2007 SBD task.