摘要
介绍了一种对ＭＰＥＧＶＩＤＥＯ码流中场景切换点进行高速检出的新方法，在计算各帧相对于预测帧的类似度之后进行了类似度的归一化，降低了ｃｕｔ检出算法的复杂度。
A new fast MPEG video scene cut detection algorithm is proposed in this paper. The similarity of a frame to it's predictive frames was calculated based on corresponding motion vector in it's MPEG stream and was further normalized. Then, a scene cut can be detected according to the normalized similarity. Test results have shown that this method greatly increased the speed and accuracy of MPEG video scene cut detection.
出处
《中国图象图形学报（A辑）》
CSCD
1999年第4期323-326,共4页
Journal of Image and Graphics
