基于二维与三维表面映射的三维面部模型拟合

Three-dimensional Face Model Fitting Based on Two- and Three-dimensional Surface Mapping

针对现有的面部模型拟合方法需要人工标记三维模型显著点,而且存在二维与三维非刚性对齐中轮廓在参数变化下消失的问题,提出一种基于二维与三维表面映射的三维面部模型拟合的方法;首先初始化参数计算人脸平均模型,然后利用Snake方法进行二维标志点和三维显著点的对齐,再更新迭代标志点和脸颊边缘,优化混合目标函数来更新形状和姿态参数,最后得到最终参数来拟合三维面部模型。结果表明,该方法可应用于单张人脸图片拟合三维人脸模型,实现三维显著点和图像中的二维标志点的自动化对齐。

In view of the fact that the existing face model fitting methods had to label the three-dimensional salient points manually,and the contour disappeared under the change of parameters in the process of two- and three-dimensional non-rigid alignment,a method of three-dimensional face model fitting based on two- and three-dimensional surface mapping was proposed.Firstly,the average face model was calculated by initializing parameters,then,Snake method was used to align two-dimensional landmarks and three-dimensional salient points.The shape and pose parameters were updated by the landmarks and cheek edge iteratively and optimizing the mixed objective function,then the three-dimensional face model was fitted by the final parameters.The results show that this method can be applied to fit a three-dimensional face model with a single face image,and realize the alignment of three-dimensional salient points and two-dimensional landmarks automatically.