摘要
文章先对压缩视频直接进行镜头分割,然后选取熵值最大的镜头进行检索,通过对视频运动进行统计,把镜头运动量化到二维概率分布空间,选择距离与样本最较近的目标,大幅缩小检测范围,然后再进行关键帧比对,实现对视频的快速准确检索。
In this paper, the shot detection in compressed domain is made first, then the shot with the largest entropy value is chosen to compare with the database. By extracting statistics of video motion and quantizing them into the form of probability distributions, the distance of motion probability distributions is compared and the ones within certain scale are chosen. Thus the scope of searching is reduced. Finally the key frames are compared, are the fast video retrieval is realized.
出处
《通信技术》
2009年第4期162-164,共3页
Communications Technology
