基于网格光的三维人脸曲面测量  

3D Facial Surface Acquisition Based on Grid Light

作　　者：石磊[1] 杨新[1] 潘海朗[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,博士上海200240

出　　处：《电子测量与仪器学报》2007年第3期9-14,共6页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金资助项目(编号:60572154)

摘　　要：本文提出了一种借助网格光从两张二维图片中测量三维人脸曲面的方法。为了保护皮肤和眼睛并且降低成本,本文中使用的网格光来自普通的白光光源。首先,一个简单的并且未编码的网格图案被投影到目标人脸上;然后,一个基于边缘信息的分水岭算法被用于提取和细化网格条纹,一个基于图形联系关系的方法被用于定位与标记网格的交叉点,利用这个网格交叉点的标记集一个迭代扩张的立体匹配算法可以执行交叉点的立体对应;最终,依据双目视觉的原理计算人脸的三维深度信息。所得的实验结果展示了我们提出的方法的可行性。In this paper, an algorithm for measuring 3D surface of human face from two 2D images using grid light is presented. The grid pattern is illuminated by incandescence light instead of laser in order to protect human eyes and skin, and reduce cost. An un-coded grid pattern is projected on the human face. Then an edge-based watershed algorithm is used to extract and thin the grid stripes. A new method based on graph connectivity is also used to locate and label the grid intersections. According to labeling principle, a labeling set is build. With the help of the labeling set, an iteratively expanding stereo matching algorithm is further used to match the intersection points. Finally, 3D depth information of the human face is calculated using binocular stereo principle. Experiment results confirm the feasibility of the proposed method.

关 键 词：人脸建模 网格光 分水岭算法 双目视觉 

分 类 号：TN247[电子电信—物理电子学]