一种基于非监督聚类的视频镜头边界检测算法
A Method of Video Shot Boundary Detection Using Unsupervised Clustering
-
摘要: 在分析归纳现有的镜头边界检测技术的基础上,提出了一种分层结构的基于非监督聚类的视频镜头边界检测算法.实验结果表明所提出的新算法不仅对切变的检测取得了较高的查全率和准确率,而且对目前技术仍不成熟的渐变检测也取得了较满意的结果.另外针对不同类型的视频,该算法的检测质量具有较强的鲁棒性.Abstract: According to the analysis of the existing shot boundary detection algorithms,a new algorithm based on moment-preserving clustering is pooposed.Simulation results indicate that the proposed algorithm can not noly improve the recall and the precision of detecting shot boundaries,but slso provide satisfactory results of gradual shot boundaries detection.In addition,the algorithm is robust to various types of video sequence.
下载: