视频检索中的镜头边界检测方法研究

Research on Shot Boundary Detection Method in Video Retrieval

视频检索要以镜头为单位进行检索,因此需要对一段完整的视频进行镜头分割。本文针对现有视频检索的镜头边界检测方法中存在的对闪光、平移和旋转等情况敏感问题,进而导致漏检和错检的不足,提出了一种基于HSV和互信息量的视频镜头边界检测算法去分割镜头,该方法在提取颜色特征的基础上能降低一定的漏检数,且选取信息学特征互信息量来对镜头进行二次边界检测,对错误的检测进行筛除,进而提高查准率和查全率。仿真实验表明,该方法比基于直方图、像素、边缘检测等方法在查准率和查全率方面有很大提高。

Video retrieval needs to be carried out in units of shots,so it is necessary to segment a complete video into shots.Aiming at the problem of sensitivity to flash,translation and rotation in the existing methods for shot boundary detection in video retrieval,which leads to the shortcomings of missed detection and error detection,this paper proposes a video shot boundary detection algorithm based on HSV and mutual information to segment shots,this method can reduce the number of missed detection to a certain extent on the basis of extracting color features,and select the mutual information of informatics features to carry out secondary boundary detection on the shots,sifting out the wrong detection,thus improving the precision and recall.Simulation results show that this method has greatly improved the precision and recall compared with histogram,pixel,edge detection and other methods.