In order to process video data efficiently, a video segmenting technique must be required. We propose a fast shot transition detecting algorithm directly on MPEG compressed video sequence. The algorithm can detect not...In order to process video data efficiently, a video segmenting technique must be required. We propose a fast shot transition detecting algorithm directly on MPEG compressed video sequence. The algorithm can detect not only abrupt transition, but also gradual transition. The computing cost of the algorithm is low, because we directly use the type of macroblocks and motion vectors that MPEG compressed video provides. The result of experiment is rather well.展开更多
基金the National Natural Science Foundation of China(60173045/F020204)
文摘In order to process video data efficiently, a video segmenting technique must be required. We propose a fast shot transition detecting algorithm directly on MPEG compressed video sequence. The algorithm can detect not only abrupt transition, but also gradual transition. The computing cost of the algorithm is low, because we directly use the type of macroblocks and motion vectors that MPEG compressed video provides. The result of experiment is rather well.
文摘提出了通过时域局部线性预测,在统一框架中准确、高效地检测视频镜头的切变和叠化等镜头变换的新方法.通过理论分析和基于视频数据的线性回归,分别得到叠化预测模式和平稳镜头预测模式.通过同时比较分析两个线性预测模式的误差,在统一框架中检测切变和叠化变换.由于直接在视频流压缩域中进行分析,因此节省了视频解码的时间.采用新方法检测时仅需要对整数进行加、减、比较和移位操作,计算复杂度很低,适合在低功耗、嵌入式设备上运行.采用MPEG-7标准测试视频的实验结果显示:切变检测查全率为95.5%,查准率为93.9%;叠化变换检测查全率为82.2%,查准率为76.0%.检测一段38 min的视频耗时仅3.5 s.