三维人脸建模中面部特征轮廓线的提取  

Facial Feature Contour Extraction for 3D Head Modeling

作　　者：高鹏东[1] 彭翔[2] 田劲东[2] 刘则毅[3] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072 [2]深圳大学光电子学研究所教育部光电子器件与系统重点实验室,广东深圳518060 [3]深圳大学理学院数学系,广东深圳518060

出　　处：《系统仿真学报》2006年第8期2105-2108,2113,共5页Journal of System Simulation

基　　金：国家自然科学基金(60275012);广东省自然科学基金(031804);广东省普通高校自然科学研究重点项目(04Z010);深圳市科技计划项目(200341)联合资助

摘　　要：提出一种基于彩色Snake模型和人脸先验知识的面部特征轮廓线提取算法,并将其用于真实感的三维人脸建模。根据肤色和灰度信息,在图像分割的基础上,利用种子填充技术获得面部各非肤色特征的轮廓初值;然后采用一种改进的彩色Snake模型对轮廓初值进行精确定位;接着,根据人脸的先验知识,利用含有方向信息的形态学边缘检测算子通过一次逐点考察完成下颚轮廓线的提取。实验结果表明该算法在轮廓线的搜索效率和定位精度上都有良好表现,在真实感的三维人脸建模以及面部识别等领域具有广泛的应用。A new approach of facial feature contour extraction for realistic 3D facial modeling is proposed, based on the color Snake model and some prior facial knowledge. On the basis of image segmentation by complexion and gray information, some rough facial feature contours can be outlined by seed-filling the non-complexional regions. Then a modified color Snake model was used to fine-draw the rough contours. Finally, based on some prior facial knowledge and directional morphological edge detector, the chin can be traced by checking the point once in turn. Experiment results indicate our approach has good performance in both speed and accuracy. Therefore it has a wide range of applications in both realistic 3D facial modeling and face recognition system.

关 键 词：彩色Snake模型 轮廓线提取 图像分割 形态学边缘检测算子 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]