摘要
为了在大规模数据下有效提升拷贝检测速度,提出了一种压缩域视频拷贝检测算法,采用2级匹配框架,利用压缩域上DCT系数顺序度量特征进行相似度匹配,显著减少了查询视频解码操作;并为DCT系数顺序度量特征建立倒排索引,以加快第一级匹配的速度。实验结果表明,与已有的算法相比,所提出算法在保持相当的检测精度条件的同时,大幅提升了检测速度。
To reduce the detection time efficiency under large scale data environment, a rapid algorithm was proposed on compressed domain using a two-level hierarchical detection scheme. The ordinal measures of DCT coefficients were adopted as visual features for similarity-matching in order to reduce the computational load in video decoding. Inverted indexing structure was used to accelerate the first level detection process. The experiment results show, compared with the previous algorithm, the algorithm can improve the detection speed obviously with the similar detection precision.
出处
《通信学报》
EI
CSCD
北大核心
2009年第3期135-140,共6页
Journal on Communications
基金
国家重点基础研究发展计划("973"计划)基金资助项目(2007CB311100)
国家高技术研究发展计划("863"计划)基金资助项目(2007AA01Z416)
国家自然科学基金资助项目(6077356)
北京市科技新星计划项目(2007B071)~~
关键词
视频拷贝检测
压缩域
DCT系数
顺序度量
video copy detection
compressed domain
DCT coefficient
ordinal measures