摘要: 在视频拷贝检测中,检测速度和检测精度不能同时达到最优。为此,提出一种基于镜头时长和纹理信息的视频拷贝检测方法。以镜头时长作为特征建立倒排索引,并放入数据库,对查询视频进行分割,将其与数据库匹配,得到初次拷贝检测结果,分析关键帧序列的纹理复杂度得到最终的拷贝检测结果。实验结果表明,该方法的检测速度较快,查准率和查全率均为1。
关键词:
视频拷贝检测,
镜头时长,
倒排索引,
纹理信息,
拷贝检测结果
Abstract: There is one serious drawback in video copy detection, which is that the high detection speed and the high detection precision can not be achieved simultaneously. To overcome this problem, this paper proposes a video copy detection method based on shot duration and texture information. It uses long lens as characteristics to establish inverted index structure, puts them into database, makes the query video segmentation, and matches them with the database to get the first copy detection results, then uses texture information of key frames to verify the detection result. Experimental results show that this method has good detection speed, and the precision rate and recall rate are 1.
Key words:
video copy detection,
shot duration,
inverted index,
texture information,
copy detection result
中图分类号:
俞芳, 崔少彬, 高振彦. 基于镜头时长和纹理信息的视频拷贝检测[J]. 计算机工程, 2013, 39(2): 304-310.
SHU Fang, CUI Shao-Ban, GAO Zhen-Pan. Video Copy Detection Based on Shot Duration and Texture Information[J]. Computer Engineering, 2013, 39(2): 304-310.