摘要
针对不同视频拷贝的相似性自动鉴别问题背景,给出了一个兼顾效率及效果的由粗到精的两阶段方案.首先抽取视频帧内容的模糊颜色直方图特征向量,构成特征空间中反映该视频性质的特征点集合和相应轨迹线,而后从特征点空间分布的角度计算位置分布直方图向量,在此基础上实现初步快速过滤,最后用简化时序特征轨迹线匹配实现更精确的相似视频鉴别.
Aiming at the problem of auto-differentiation of similar video clips, a coarse to fine two phase method is presented, which take into account both effect and efficiency. Firstly, Feature point set and corresponding trajectory are obtained based on fuzzy color histogram feature vector. Secondly, Position density histogram in the feature space of the video clip is computed and elementary quick filtration is implemented based on this statistical feature. Finally, More accurate differentiation is performed using simplified time sequence trajectory after the elementary quick filtration.
出处
《小型微型计算机系统》
CSCD
北大核心
2007年第8期1477-1481,共5页
Journal of Chinese Computer Systems
关键词
视频
特征点
轨迹线
相似性
video
feature point
trajectory
similarity