摘要： 镜头边界检测是基于内容视频检索的重要组成部分。为从不同类型的视频中有效地检测出视频镜头边界，提出一种视频镜头边界检测算法。通过视频帧图像的颜色特征，得到视频的相似性矩阵，根据突变镜头和渐变镜头在Affinity Propagation聚类结果中的不同特点，运用双阈值法检测镜头边界。实验结果表明，该算法从视频的本身信息分布出发，能自动快速地检测出镜头边界。

关键词: 视频检索, 镜头边界, 视频镜头, 聚类

Abstract: Shot boundary detection is important component of content-based video indexing. For detecting shot boundary efficiently from different video, this paper presents an efficient method for shot boundary detection. It gets the shot comparability matrix by histogram differences, according to the different characteristics of abrupt change and gradual change shot in Affinity Propagation clustering results, detect shot change by twin-comparison. Experiment results show that the algorithm can detect shot boundary automatically and fast from video information distribution.

Key words: video indexing, shot boundary, video shot, clustering

中图分类号: 

• TP391