摘要
对基于内容的视频检索中视频镜头检测技术进行研究。首先,特征的选取,使用符合人类感知的HSV空间,并以该空间上的非均匀分块加权颜色直方图作为帧间差特征;其次,以双阈值比较法为检测方法,对其存在的问题进行改进;最后,采用基于局部统计特性的阈值选取方法。实验表明:本方法能取得较好效果,在一定程度上提高检测的查全率和准确率。
A video shot detection based on CBVR is presented. The weighting color histogram of divided no nuniformly in HSV as the difference characteristic among the frames is adopted to improved twain - comparison as detection method, and based on practical statistics characteristic as threshold chosen. The experiment shows that the method can be got better resule, which can improve the recall and precision.
出处
《计算机与数字工程》
2008年第5期164-166,共3页
Computer & Digital Engineering
