三维人脸的非均匀重采样对齐算法

    Nonuniform Resampling Based on Method for Pixel-wise Correspondence Between 3D Faces

    • 摘要: 为了用尽量少的网格点和面片数表示对齐后的三维人脸,基于分片的网格重采样思想,通过分析每片的曲率高低自动确定各片的网格稠密度,得到非均匀的网格结构,实现了不同三维人脸数据间点到点的一一对齐.实验结果表明:与以往的对齐算法相比,此方法不仅可以大大减少表示人脸的网格点数目,还可以较好地保持人脸的形状和外观特征;使用非均匀重采样做对齐后的三维人脸库建立的人脸模型进行人脸重建,比使用均匀重采样得到的人脸模型在重建速度上有了很大提高.

       

      Abstract: In order to get few vetices and 3D faces, patche-based method is used for gridded-resampling, which decides the sampling density automatically over the face patches by analyzing each patch's curvature. The method produces nonuniform mesh and achieves the point-to-point correspondence of different 3D faces. The experimental results show that the method is not only able to reduce the number of mesh vetices of 3D face after resampling, but it can also keep the similarity of geometry and appearance well. And the performance of the face model constructed by nonuniform resampling is higher than that constructed by uniform resampling.

       

    /

    返回文章
    返回