基于Level Set方法的人脸轮廓提取  

Face contour extraction using level sets

作　　者：柳长青[1] 田振夫[1] 

机构地区：[1]宁夏大学数学计算机学院,宁夏银川750021

出　　处：《计算机工程与设计》2008年第12期3179-3181,3184,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(10662006);宁夏大学科学研究基金项目(ZR0628);2007年度宁夏高校科学研究基金项目

摘　　要：人脸轮廓提取在诸如安全检查、保安管理、交通电子监控、机器人研究等方面有着十分重要的应用价值。先利用图像背景差分法快速检测出人脸的初始矩形轮廓线,然后采用LevelSet方法对视频图像进行人脸轮廓提取,Level Set演化函数空间方向上采用四阶紧致差分逼近式离散,计算过程中加入了窄带算法和全局优化方法。数值实验结果表明,该算法在不增加计算时间的基础上,可以检测出模糊或离散的边界,并得到精确的人脸轮廓线。The extraction of face shape has significant application in many fields such as safety check, management of ensuing public security, electronic monitoring in traffic and research in robot etc, Firstly, the initial rectangle of face is detected by using image background difference quickly. Face contour is extracted by using level set method from video image. Secondly, level set evolutionary function is descritized by the fourth-order symmetrical compact finite difference scheme in spatial direction. Narrow-band algorithm and global optimization methods are adopted in computation. Finally, the numerical experiments are carried out to show that the presented method has many advantages in detecting both vague and discrete border and extract more accurate face contour without increasing computational time. Additionally, it can also satisfy the demand of real-time extraction. Based on investigations above, a conclusion is drawn that level set method is effective and can be applied to extracting face shape.

关 键 词：LevelSet方法 轮廓提取 紧致差分 计算机视觉 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]