摘要
关键帧提取技术是基于内容视频检索的基础。针对压缩视频序列提出一种基于镜头的关键帧提取改进算法。其核心是根据压缩视频序列的编码特点,只需要对其进行部分解码,利用I帧信息的直流分量信息构造DC缩微图,并结合图像帧不同区域DC信息的重要程度差异进行相似性度量,进而实现关键帧提取。实验证明,该算法较之传统算法,在查全率和检索时间两项指标上均有显著改善,尤其适用于新闻纪录片、电影片等局部运动较为剧烈的视频序列。
Key frame extraction is one of fundamental processes in content based on video retrieval （CBVR）. This paper proposed an improved key frame extraction algorithm based on shot for MPEG compressed video, which was based on different formulas comparing discrete cosine transform （DCT） direct current （DC） eodfficients over the I-frames in MPEG video stream, and for which only minimal decoding was needed. In particular, introduced the importance difference of DC codfficients in one image frame into the process of similarity measure. Experimental results show that our algorithm is superior to other traditonal ones in two criterions： recall and retrieval time, and it is a suitable method for newsreel or film, which have severe local motions in video sequences.
出处
《计算机应用研究》
CSCD
北大核心
2009年第9期3539-3541,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60872081)
北京交通大学人才基金资助项目(W07J0260)
关键词
关键帧提取
压缩域
MPEG
key frame extraction
compressed field
MPEG
