基于累积帧的自适应双阈值镜头边界检测算法

摘要

镜头边界检测是基于内容视频检索的重要组成部分.针对渐变的视频序列,提出一种基于累积帧的自适应阈值镜头边界检测方法.考虑用一个累积帧把每一帧之间的差异累积起来,放大渐变的特征变化,利用块匹配算法对运动进行估计和补偿,以减少物体运动对镜头边界检测的影响.通过分析帧亮度和饱和度与阈值的关系,用自适应双周值提高镜头边界检测的准确性.实验结果表明,该算法可以在较小的计算强度下较好地实现对镜头渐变的检测.%Improving the precision of shot boundary detection is very important. This paper presented an algorithm for shot boundary detection based on self-adaptive dual thresholds of accumulative frame. It used an accumulative frame to memorize the accumulative differences among the sequential frames of the video and magnify gradual feature changes. A block-matching algorithm was designed to compensate for the motion of the object to reduce the effects of it. Self-adaptive dual thresholds were adopted to improve the veracity of the shot boundary detection. The proposed method can well detect gradual shot with less calculating strength by experiment on various video clips.