一种适合于监控视频内容检索的关键帧提取新方法-《郑州大学学报(工学版)》

文章信息/Info

Title:: A Novel Method of Key Frame Extraction for the Surveillance Video Content Retrieval

Author(s):: ZHOU Bing1; HA0 Wei-wei1; YUAN She-feng2; 1.Sehool of lnformation Engineering , Zhengzhou University , Zhengahou 450001, China; 2.School of Electronic Information En-gineering，Henan Vocational College of Agriculture，Zhengzhou 451450 , China

摘要:: 为减少监控视频的存储容量,便于用户迅速查看视频关键信息,提出一种适于监控视频内容检索的关键帧提取方法.首先,使用背景差分法检测含有运动物体的关键视频段,记录其起始帧和结束帧,然后,在关键视频段中基于联合直方图的方法进行关键帧的粗提取,最后,根据监控视频帧序列的连续性特征,通过图像的信息熵进一步精确提取关键帧,实验结果表明,该方法能有效地提取出监控视频中的关键帧,大大减少了视频数据的存储量,且便于用户浏览关键信息.

Abstract:: This paper proposes a new method of the key frame extraction aiming at reducing storage and fastretrieval of the surveillance video.First,an improved background subtraction is used to detect the key videosegment which contains the moving objects,and record the first frame and the last frame of the segment.Thenthe key frames are extracted coarsely in the video segment based on the joint histogram. Finally,according tothe continuity of the key video frames,we further refine the key frames by image information entropy. The ex-perimental results show that the method can extract the key frames of the surveillance video effectively.Itgreatly reduces the storage of the video data,and helps the users to browse key information easily.