摘要
在已有的三维人脸重建方法的基础上,提出一种使用姿态估计进行深度信息恢复,并用于从单幅图像重建三维人脸模型的方法.首先将已检测到的特征点和通用三维人脸模型进行配准得到人脸姿态参数,再将姿态参数作用在通用三维人脸模型上得到特征点的深度信息;接着利用特征点及相应的深度信息对通用三维人脸模型进行全局变形和局部变形,生成人脸几何形状,并采用保相似的网格参数化方法计算纹理坐标,得到最终的三维人脸模型.实验结果表明,此方法在提高重建效率的同时保持了较好的视觉效果.
On the basis of the existing 3D face reconstruction method,using pose estimation for depth information recovery was proposed,which is used for 3D face reconstruction based on a single image.It first obtained face pose parameters by matching the detected feature points to a general 3D face model,then applied pose parameters to the generic 3D face model to get depth information.Secondly,global and local deformation were made by detected feature points and their corresponding depth information to get the face geometry,then obtained texture coordinates by an as-similar-as-possible planar parameterization,finally got the 3D face model.Experiments have shown that using depth information recovered from pose estimation was efficient and easy to achieve good results.
出处
《微电子学与计算机》
CSCD
北大核心
2015年第9期95-99,104,共6页
Microelectronics & Computer
基金
北京市自然科学基金资助项目(4132068)
关键词
人脸三维重建
姿态估计
深度信息恢复
单幅图像
3D face reconstruction
pose estimation
depth information recovery
single image
