基于形变模型和随机照片集的人脸三维重建

Face 3D Reconstruction Method Based on Morphable Model and Random Photo datesets

目前三维人脸重建方法很难得到人脸细节,较难适应于随机照片集,针对此问题,提出一种可以适应于无约束照片集的人脸重建方法。首先,通过联合特征点,采用三维形变模型(3DMM)来生成个性化样本,在联合朗勃照片渲染公式中解决了光度立体;然后,提出一种关联性测量来权重人脸各个部分的影响,在法线估计和表面重建中交替进行由粗到精的处理,得到人脸三维模型。实验使用合成数据和真实数据集进行性能测试。结果表明,所提方法可以重建详细的个人三维人脸模型,模型精度比其他方法更加精确;并在一定程度上提高了计算效率,可适用于无约束的照片集。

3 D face reconstruction method is difficult to get face details and hard to adapt to the random photo datesets.To solve this problem,a face reconstruction method which can adapt to the unconstrained photo set is proposed in this paper.Firstly,by combining with the feature points,the 3D morphable model(3DMM)is used to generate the personalized sample,and the photometric stereo is solved in the joint Lambert photo rendering formula.Then,the correlation measurement is proposed to weigh the influence of each part of the face,and the processing is made alternately from the coarse to the fine in the normal estimation and surface reconstruction to get the 3D face model.The performance tests are conducted by using synthetic data and real datasets.The results show that the proposed method can reconstruct the detailed personal 3 D face model,and the accuracy of the model is higher than other methods.Meanwhile,the computational efficiency is improved to a certain extent,and the method can be applied to unconstrained random photo datesets.