基于背景颜色不变量的改进双直方图镜头边界探测算法  

An Improved Twin-Histogram Algorithm for Shot Detection Based on Background-Color-Invariable

作　　者：雷震[1] 吴玲达[1] 老松杨[1] 

机构地区：[1]国防科技大学多媒体研究开发中心,长沙410073

出　　处：《计算机工程与应用》2003年第29期35-37,115,共4页Computer Engineering and Applications

基　　金：国家863高技术研究发展计划项目基金资助

摘　　要：快速运动一般不是发生在背景内,而是发生在对象内,因此关注背景区域对于探测镜头边界是行之有效的,而在进行镜头边界探测时,普通双直方图法对渐变边界检测效果比较好,而对突变镜头检测的效果不是很理想。提出了一种基于背景颜色不变量的改进双直方图算法进行镜头边界探测,该方法首先定义出背景区域,然后选用Lab彩色模型来计算背景区域的色度,再利用改进的双直方图算法进行镜头边界探测,该方法所需计算量较小,且可保证较高的精度。实验结果表明,该算法的分割效果明显优于其他几种算法,具有更高的探准率。Rapid movement often happens in the objects but not background,so attention to background areas is feasible for shot detection,whereas common twin-histogram algorithm only fit for detecting gradual transition but not fit for detecting abrupt transition.In this paper,an improved twin-histogram algorithm based on background-color-invariable is presented to detect shot boundary.Firstly,the background area is defined.Secondly,L  a  b  color model is used to calculate the Hue in the background areas,Lastly,the improved twin -histogram algorithm is used to detect shot boundary.This method needs a few calculations and has higher precision ratio.The experimental results show that the method is more accurate and effective in shot detection than other algorithms.

关 键 词：镜头边界探测 色度直方图 双直方图法 突变模型和闪光灯模型探测器 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP301.6[自动化与计算机技术—计算机科学与技术]