摘要
文提出了一种基于H.264/AVC视频流的镜头边界检测算法,利用宏块类型信息分析视频帧之间的时间相关性和空间相关性来检测镜头突变,并通过分析宏块类型信息选取出渐变候选帧,对候选帧进一步考察前景及后景的平均运动活性度以及其后面连续P帧的帧内编码宏块数的方法检测镜头渐变,此方法不需对压缩视频进行解压,计算量小,实验证明,对于运动强度较大的视频流也有较好效果,并有效克服了以往算法中对突发的轻微局部运动或全局运动不敏感的缺点。
The paper proposed a shot-change detection algorithm for H.264/AVC compressed video.It detect the abrupt transitions by analyzing temporal dependencies and spatial similarities between neighboring frames.By analyzing the number of intra-coded macroblocks of the P fanmes which behand the candidate gradual transition frames and the mean motion intensitys in both foreground and background can detect gradual transitions.The computational complexity of this method is lower,because it is not need to decode the video stream,the experimental results shows that the proposal algorithm is also efficient for the high motion intensity video streams and video streams which contain the sudden changes in light intensity.
出处
《微计算机信息》
2010年第30期173-175,共3页
Control & Automation
