含有复杂运动的MPEG压缩域镜头分割算法

A Compressed Video Partition Algorithm of MPEG Including Complex Motion

本文针对包含有复杂运动的MPEG视频流,给出了一种压缩域的镜头分割算法。该方法主要应用DC亮度系数直方图差变化、I图密度分布、B图的双向运动向量数比的变化及相应的曲线振荡频度,获得镜头分割,并进而可检测出切变、渐变及复杂运动等。该算法已在Pentium II PC机上实现,并与传统方法进行了比较,得到较满意的结果。实验表明该方法较适用于含有复杂运动的视频流的镜头分割,不仅可以应用于新闻视频,而且对含有较大运动及较多特技的视频流,有较好的镜头分割效果。

Focusing on MPEG stream including complex shot motion, a shot partition algorithm in compressed domain is presented. In the proposed algorithm, following factors are used to determine the shot partition: changes on DC luminance coefficient histogram difference, the density distribution of I pictures, the variation of the ratio of the number of motion vectors between in forward and backward directions in B pictures, and the relative curve vibration frequency. Finally, features such as shot cut, gradual transition and complex motion, etc can be extracted. This algorithm of partition not only beneficial to complex shot motion video streams such as news video but also to those including fast motion or special effect.