| Summary: | 碩士 === 國立中正大學 === 資訊工程所 === 93 === In this study, a two-stage segmentation scheme for H.264 video is proposed. The objective of the proposed scheme is to detect video transitions more efficiently for H.264 video. For general video sequences, they are usually composed of many shots, and it must exist a video transition between one shot and the next shot following it for all shots. A video transition only can tells viewers a shot will change into another shot and it is no meaning for contents of a video sequence. Therefore, before video indexing, video summarizing, or video retrieval, it must finds video transitions and segments a video into many shots without video transition parts. In the first stage of the proposed scheme, a frame-based macroblock modes statistics method is used to detect possible video transitions zones quickly in H.264 video. In the second stage, a histogram-based luma and chroma comparison method are used to verify whether video transitions are existing or not in possible video transitions zones. If video transitions indeed exist, then finding the certain frame numbers of video transitions in possible video transitions zones. Avoiding high complex computing and comparison frame by frame, the proposed two-stage segmentation scheme can make video transitions detection more efficiently for H.264 video. The proposed algorithm usually has the best performance (the recall and the precision) in most simulation cases. This shows the feasibility of the proposed approach.
|
|---|
