一个自动阈值选择的镜头检测算法  

Novel Shot Detection by Automated Threshold Selection

作　　者：原野[1] 宋擒豹[1] 沈钧毅[1] 倪冰[1] 

机构地区：[1]西安交通大学计算机系,陕西西安710049

出　　处：《小型微型计算机系统》2004年第7期1337-1340,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金 (60 173 0 5 8)资助

摘　　要：对现有视频分割工作进行了简要回顾 ,并在已有的 Twin Comparison方法基础上 ,提出了一种新的象素域镜头检测算法 .该算法使用了一个滑动窗口进行镜头边界识别 ,仅需要一趟视频解码和扫描 ,提高了算法的速度 ,这样该算法可以应用于实时视频流的镜头检测 .同时该算法够根据滑动窗口中的帧间颜色直方图差异值来动态计算用于识别镜头变换的局部阈值 ,提高了镜头边界检测算法的精度 .实验结果显示 ,文中所述方法在速度和精确度上都优于This paper reviews the shot detection methods briefly. And a novel algorithm for shot detection is proposed based on the improvement to the Twin Comparison algorithm. The new algorithm uses a sliding window to identify the boundaries of shots and just need one pass of video decoding and scanning. So the speed of the algorithm is improved. And the algorithm is applicable to detect shots in real time video stream. The local thresholds to identify the shot transitions are computed dynamically using the color histogram difference between consecutive frames in the sliding window. So the precision of shot detection is improved. The experimental results show that the new algorithm outperforms the Twin Comparison in speed and precision.

关 键 词：视频分割 镜头检测 颜色直方图 阈值选择 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]