改进的基于主成分分析的三维人脸形状建模

针对基于主成分分析(PCA)的三维(3D)人脸形状重构不精确的缺点,对位移向量、缩放因子、旋转矩阵、重构系数等参数进行了精确的定义和计算,并对重构模型进行了有效的调整。根据CVL Face Database建立了113个3D人脸形状模型,经迭代运算后重构出了由特征点组成的稀疏3D人脸形状模型。利用径向基函数(RBF)得到了密集3D人脸形状模型。试验结果表明,算法可以得到精确的重构模型,对深度旋转的人脸图像和特征点定位误差也有很好的鲁棒性。

Improved three dimensions face shape reconstruction based on principal component analysis

Against the imprecision of the reconstructed three dimensions(3D)face shape model by principal component analysis(PCA),the translation vector,scale factor,rotation matrix,coefficients of the shape eigenvectors are defined and com- puted,and the effective adjustment to the reconstructed 3D face shape modal is carried on.The 113 3D face shape models are established based on CVL face database.The spare 3D face shape rhode1 composed of feature points is reconstructed.The final 3D face shape model is obtained using radius base function(RBF).Experiment results show that the proposed method can ob- tain more precise reconstruction model,and more robust with face image under depth rotation and feature points position error.