基于共形几何的面部模型插值构建算法

Facial model interpolation reconstruction algorithm based on conformal geometry

针对传统三维模型构建方法难以获取连续维度的高质量面部模型问题,提出一种以不同表情的多面部模型为数据源基于共形几何的面部模型插值构建技术。利用准共形映射算法将面部模型降维到二维平面;提出一种插值算法对多模型形状特征进行线性插值,为构建过渡模型定义几何约束,同时提出一种自适应模型选择算法在数据源中自适应选择基础模型;通过几何约束引导准共形迭代改变基础模型形状;将变化后的二维模型转化到三维空间,构建具有过渡表情的三维面部模型。实验结果表明,该算法的有效性和运行效率都得到大幅提高。

Aiming at the difficulty of obtaining high-quality facial models with continuous dimensional using traditional 3D model construction methods,a facial model interpolation construction technology based on conformal geometry was proposed using multiple facial models with different expressions as data sources.The quasi-conformal mapping was used to reduce the dimensionality of the facial model and convert it into a 2D model.An interpolation algorithm was proposed to linearly interpolate the shape features of multiple models to define geometric constraints for constructing transition models.An adaptive model selection algorithm was proposed to adaptively select the basic model among multiple facial models.The quasi-conformal iteration was guided by geometric constraints to change the two-dimensional plane shape of the basic model.The changed 2D facial model was transformed into 3D space,and a 3D facial model with transitional expressions was constructed.Experimental results show that the effectiveness and efficiency of the algorithm are greatly improved.