基于可变窗的镜头边界检测算法

A Video Shot Boundary Detection Method Based on Variable-Size Window

为了解决传统镜头边界检测算法的漏检和误检问题,对判决方法进行研究,提出了可变窗概念.当一个镜头中画面变化激烈时,减少窗口跨度;在平缓画面的镜头通过渐变过渡到另一个镜头时,增加窗口跨度.该算法统计各帧的分块直方图作为特征,结合可变窗判别机制,对非压缩域中的视频帧进行镜头边界检测.实验证明,该算法提高了整体镜头查全率,同时也降低了由于镜头画面激烈而发生的误检率.

This article introducd the variable-size window to solve the false and leak detections in traditional shot boundary detection methods. It can reduce its size when there is fierce change in the frame within a shot and increase its size when the one shot transforms into another shot through gradual transform. This method calculates the sub-frame histogram in every uncompressed frame and use variable-size window to detect the shot boundary. The experiments verify that this method can reduce leak detections and false detections caused by camera motion effectively and thus gain the general performance.