摘要
慢镜头探测是基于内容视频分析的一项重要技术。在介绍视频压缩的基本理论及MPEG-1视频压缩标准的基础上,对压缩域内的慢镜头探测进行了研究,提出了一种两阶段压缩域下的慢镜头探测方法。该探测方法的第一步是基于宏块类型信息的镜头探测,在镜头探测的基础上定义和计算镜头密度,探测可能的慢镜头出现位置;第二步是计算压缩域下连续帧间差,从候选慢镜头中探测慢镜头。实验结果显示,相对于已有的方法,该方法具有较快的探测速度和较高的准确率。
The Slow-motion replay（SMR） detection is an important technique for content-based video analysis. This paper introduced the basic theory of video compression and MPEG-1 video compression standard. A two-step method for SMR detection on MPEG uncompressed domain was proposed, in which firstly shot density was defined and calculated based on shot detection results using MPGE macroblock types information to detect the occurrences of possible SMR shots,and secondly the difference between consecutive frames was measured to detect true SMR in candidate SMR shots. Experiments results show that unlike previous approaches,our method has great improvement in both speed and accuracy.
出处
《计算机科学》
CSCD
北大核心
2009年第9期283-286,共4页
Computer Science
基金
国家自然科学基金(60572137)
国家高技术研究发展计划(863计划)(2006AA01Z316)
教育部博士点基金资助
关键词
慢镜头探测
镜头切割
基于内容视频检索
MPEG压缩域
宏块类型
Slow-motion replay detection, Shot cut, Content-based video analysis, MPEG compressed domain, Macrob- lock type
