摘要
综合利用MPEG视频流压缩域中P帧、B帧的DC系数和运动特征以及非压缩域纹理特征,提出一种基于Rough Set与SVM的关键帧提取方法。该方法首先提取视频流中P帧、B帧的DC系数、分析运动活力性和运动空间分布和宏块类型、提取帧图的纹理特征后,利用Rough Set对这些特征进行约减后,用SVM进行分类识别。实验表明,该方法能有效地识别关键帧。
An approach for key frame extraction based on Rough Set and SVM is introduced in the paper,which comprehensively utilises the DC coefficient of P frame and B frame and the motion features in compressed domain of MPEG video stream as well as the non-compressed domain texture features.First,the method extracts the DC coefficient of P frame and B frame from video stream,and then it analyses the motion-activity,the motion-space distribution and the type of macro blocks.After the frame texture features are extracted,Rough Set is employed to conduct the reduction on them and SVM is used to sort and identify them afterwards.Experimental results show that the approach is effective in key frame identification.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第6期244-249,共6页
Computer Applications and Software