摘要
针对大规模的视频版权保护中检索准确率低和检索速度慢的问题,提出一种基于球哈希和改进的灰度差算法的视频拷贝检测方法。首先,对传统的灰度差提取关键帧的算法进行改进,提出利用灰度差与灰度差累积量相结合的方法提取关键帧,减少了提取关键帧时产生的累积误差,提高视频检索的准确率;然后,提取关键帧的SIFT特征,并且利用著名的球哈希算法对SIFT特征建立索引,把128维的高维浮点向量变为二值索引,提高检索速度。实验结果表明该方法不仅提高了视频检索的准确率和检索速度,而且保证了较高的查全率。
For the problem of low retrieval accuracy and slow retrieval speed in large-scale video copyright protection, we proposed a video copy detection method, it is based on spherical Hashing and improved gray difference algorithm. First, we improved the traditional algorithm of extracting key frames based on gray difference, proposed a method to extract key frames which combines the gray difference with gray differ- ence accumulation, thus reduced the cumulative errors when extracting key frames, as well as improved the accuracy of video retrieval. Sec- ondly, we extracted the SIFT feature of key frames and used the famous spherical Hashing algorithm to establish hashing index for SIFT fea- ture, aI^d converted the 128 high-dimensional floating-point vectors to binary hashing indexes, therefore improved retrieval speed. Experimen- tal results showed that this method not only improved video retrieval accuracy and retrieval speed, but atso ensured a higher recall rate.
出处
《计算机应用与软件》
CSCD
2016年第1期181-184,189,共5页
Computer Applications and Software
基金
国家科技支撑项目子课题(2012BAH04F02-2)
山西省国际科技合作项目(2012081034-2)
人社部留学人员科技活动项目(2011-508)
关键词
版权保护
视频拷贝检测
灰度差
SIFT特征
球哈希
Copyright protection Video copy detection Gray difference SIFT feature Spherical Hashing
