压缩视频流中主要运动的定性分析

随着视频数据量的急剧增加，如何为海量视频数据建立高效索引并提供查询和检索能力成为目前信息检索领域重要的研究课题，许多研究表明，通过对视频中主要运动的定性分析，有助于实现视频内容的索引和检索，该文根据运动矢量的分布特性，运动矢量之间的相对关系特性等，提出了一种压缩视频流中主要运动的定性分析算法，实验结果表明该算法具有速度快，精度高，鲁棒性好等特性，非常适合于视频信息检索方面的应用。

Qualitative Analysis of Dominant Motion for Compressed Video Stream

Content-based indexing of video has received great attention by the research community. In general, by analyzing color, shape, texture and motion features in video data, we can get useful meta-data about video content to realize video retrieval. A new algorithm for qualitative analysis of dominant motion in video shots is proposed in this paper. It makes use of the directional distribution of motion vectors and the relationship between any two motion vectors in MPEG-1/2 stream. It is possible to obtain reliable estimation of dominant motion by directly processing motion vectors easily obtained from the compressed stream. Working directly with the compressed video greatly reduces the processing time and enhances storage efficiency. As an illustration of this idea, the algorithm is tested on the Open Video Test Data Collection. Experimental results show that the proposed algorithm is much fast, robust and of high accuracy.