真实感人脸的形状与表情空间

提出一种建立三维人脸扫描模型参数空间的算法,其中的模板拟和算法基于能量最小的优化机制,通过非线性优化过程求解模板模型在每个顶点上的位移矢量,使之逼近目标模型.优化目标方程由以下测度组成：距离、平滑度以及人脸特征对应,如特征曲线、边界和特征点对等.使用该算法可用于不同人脸以及不同表情模型之间的对应建立,从而获取一致参数化的人脸形状和表情空间.在文中系统中,三维面部特征曲线通过Canny边检测算法自动获取,这样自动检测获取的特征曲线可用于降低三维形状描述的维数,同时完整的面部几何形状通过径向基函数插值得到.在中性人脸和表情人脸模型上所作的一致参数化为许多应用提供了平台,如形状渐变,纹理迁移和表情迁移.考虑到自动提取的特征曲线和二维线画卡通人脸的相似性,使用迭代优化算法实现二维线画卡通人脸姿态到三维真实人脸模型的迁移.

Shape and Expression Space of Realistic Human Faces

We present a novel method to generate a parameterized space of human facial scans. Our approach is under an energy minimization frame. The optimization solves the displacement vector at each vertex on the base mesh. The error metrics include the following terms: distance, surface smoothness, and feature constraints including curves, boundaries and marker pairs. With the template fitting to static expression scans with the same algorithm, an expression space is generated as well. The feature curves are automatically extracted from expression maps with an edge detection algorithm. Such feature curves are used to reduce the dimensionality of the shape space. The consistent parameterization in neutral faces and expressions forms the basis of a large category of applications, e.g. shape morphing, texture transferring and expression transferring. In view of the similarity between feature curves on depth images and those on 2D line drawings of cartoons, the nonlinear optimization is employed to realize the dramatic expression transfer from cartoons onto realistic facial models.