虚拟广告系统中视频运动分割算法的研究  

Video Segmentation Algorithm in Virtual Advertisement System

作　　者：陈临强[1] 翟小瑞[1] 

机构地区：[1]杭州电子科技大学图形图像研究所,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2006年第6期1-4,共4页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：浙江省科技计划项目(2006C23054)

摘　　要：在虚拟广告系统中,为了使虚拟图像的插入具有真实感,有效的视频分割便成了其中的关键的技术。该文提出了一种背景剪除与亮度改变相结合的视频运动对象分割算法。该算法首先采用人机交互的方式,指定视频场景中需要处理的区域,取得背景帧,经RGB色彩差分处理,分割出前景物体;根据亮度变化的准则,检测出阴影区域,为后继的虚拟图像的真实感插入做好预处理准备。实验结果表明,该算法简单实用,能够实时有效地分割出视频序列中前景物体和阴影区域。In order to achieve the target of making the virtual objects appearing as real entities in the scene, video segmentation becomes to be the key technique in virtual advertisement system. A video segmentation algorithm based on background subtraction and luminance change is presented in this paper. Using interactive method, the area where we＇ll insert the virtual objects in is destined, and get the corresponded background frame at first in the algorithm. Then, the rough region we interested is found by using background RGB subtraction method. At last, based on luminance change rule, every pixel in the region is processed to estimate whether it is shadow or not. Experiment results show that the algorithm is easy to achieve and robust to differentiate real moving objects and shadow area.

关 键 词：运动分割 色彩背景差分 阴影检测 虚拟广告系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]