基于动态背景构造的视频运动对象自动分割  

Automatic Segmentation of Moving Objects in Video Sequences Based on Dynamic Background Construction

作　　者：吴思[1] 林守勋[2] 张勇东[2] 

机构地区：[1]暨南大学计算机科学系,广州510632 [2]中国科学院计算技术研究所数字化技术研究室,北京100080

出　　处：《计算机学报》2005年第8期1386-1392,共7页Chinese Journal of Computers

基　　金：北京市科技计划项目(E0004024040231);国家自然科学基金项目(60302028;60473002);北京市自然科学基金重点项目(4051004)资助

摘　　要：提出了一种基于动态背景构造的视频运动对象自动分割算法.首先,基于前景分离的动态背景构造技术使用与当前帧相邻的前后多帧图像中的背景信息准确地构造当前背景;然后,通过背景消除分割出运动对象,同时对静态前景区域(即帧间静止的运动对象区域)进行检测并将其合并到已分割出的对象区域上,从而获得完整的对象区域;最后,以对象区域的边缘为初始位置,使用以彩色梯度为外部能量的活动轮廓(snake)算法获得精确的对象轮廓.实验结果表明,该文算法有效地克服了显露背景和对象的不规则运动对分割准确度的影响,能够准确地实现视频运动对象的自动分割.In this paper, an automatic segmentation algorithm for moving objects in video sequences based on dynamic background construction is proposed. First, based on a coarse separation of foreground from images, the dynamic background construction technique, an accurate background for current frame with information in temporally adjacent frames is constructed.Then, the intact moving objects＇ areas are obtained by subtracting the constructed background from the current frame and merging the static foreground areas （namely, parts of objects stopping moving between adjacent frames） detected using the temporal information. Finally, the accurate objects＇ contours are extracted by active contour （snake） using color gradient as its external energy. Experiments demonstrate that the proposed method is robust to the influence of uncovered background and objects＇ irregular motion, and able to extract moving objects successfully from video sequences with dynamic background.

关 键 词：对象分割 动态背景构造 彩色梯度 活动轮廓 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]