摘要
针对三维人脸表情合成精度较低的问题,提出一种基于NURBS曲面拟合技术的三维人脸表情合成算法.基于NURBS曲面的优化控制原理,将二维人脸图像转化为连续的三维表情参考模型,通过增加整体几何约束和局部平滑约束,提高了人脸表情形状的平滑程度和嘴部拟合的精确度,生成了更加精细的三维人脸表情形状.仿真结果表明,该算法对不同环境、姿势和光照条件的二维人脸图像,具有一定的稳定性和泛化性,合成效果比较理想.
In order to solve the problem of low accuracy for 3D facial expression synthesis,a 3D facial expression synthesis algorithm based on NURBS surface fitting technology was proposed.Based on the optimal control principle of NURBS surface,a 2D facial image was transformed into a continuous 3D expression reference model.By adding global geometric constraints and local smooth constraints,the smoothness of facial expression shape and the accuracy of mouth fitting were improved to generate a more refined 3D facial expression shape.The simulation results show that the as-proposed algorithm shows particular stability and generalization for 2D facial images under different environments,poses and lighting conditions,and the synthesis effect is quite ideal.
作者
伍菲
WU Fei(School of Information Science and Technology, Guilin University of Electronic Technology, Guilin 541004, China)
出处
《沈阳工业大学学报》
CAS
北大核心
2021年第6期688-691,共4页
Journal of Shenyang University of Technology
基金
广西教育厅高校中青年教师基础能力提升项目(2017KY1341)
广西高等学校千名中青年骨干教师培育计划项目(2020QGRW037).
关键词
NURBS曲面
表情合成
特征点
人脸轮廓
二阶偏导
鲁棒性
泛化性
平滑约束
NURBS surface
expression synthesis
feature point
facial contour
second partial derivative
robustness
generalization
smooth constraint
