基于曲率信息的三维人脸面部特征提取  

3D facial feature extraction based on curvature information

作　　者：郭哲[1] 张艳宁[1] 林增刚[1] 

机构地区：[1]西北工业大学计算机学院,西安710072

出　　处：《计算机工程与应用》2007年第24期239-241,245,共4页Computer Engineering and Applications

基　　金：西北工业大学研究生创业种子基金资助项目(No.Z200654)

摘　　要：提出了一种基于曲率信息的三维人脸面部特征提取方法。首先根据几何约束和曲率信息得到由若干间断点描述的人脸五官区域轮廓线;其次采用边界连接及变形膨胀方法,得到完全闭合、但存在多余毛刺的区域边界;最后利用二维图像处理方法进行边界精确化处理,提取出比较精确的人脸面部特征区域。实验证明了所提方法对正面自然表情三维人脸模型的有效性。A new method of 3D facial feature extraction based on curvature information is proposed in this paper.First,extracts facial feature contour lines which is composed with some discontinuous points based on curvature information and geometrical restriction.Second,obtains closed but including burrs contour lines by using contour lines connection and deformed dilation.At last, gets precise facial feature regions by using 2D image processing method.Experiments prove that the proposed method is effective to frontal and natural expression 3D modles.

关 键 词：面部特征曲率信息边界连接变形膨胀 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]