一种三维人脸点云数据的二维映射表示

A Two Dimensional Mapping Representation of 3D Human Face Point Cloud

摘要: 针对三维人脸点云存储和计算成本过高的问题，提出了一种基于人脸点云球面拟合和球面映射的三维人脸点云数据的二维表示方法。利用原始人脸点云的球面拟合确定以该人脸点云为中心的球面坐标系，通过球面投影和邻域插值分别展开为二维球面深度和纹理图像。这种二维表示法没有信息丢失，可以完整恢复原始三维信息。实验表明提出的二维形式表示的人脸图像具有更低的存储和处理计算消耗，并保持了原始三维人脸的鉴别信息。该三维人脸的二维表示法对三维人脸海量存储和识别具有实际意义。

Abstract: To solve the problem of high cost of the storage and processing of 3D human face point cloud, a two-dimensional representation method of 3D human face point cloud based on spherical fitting and spherical mapping is proposed. The facial point cloud spherical fitting of the original face is used to determine the spherical coordinate system centered on the point cloud of the human face. The 2D spherical depth image and texture image are respectively mapped by spherical projection and neighborhood interpolation. The proposed representation has no information loss and can completely restore the original three-dimensional information. Experiments show that the proposed 2D face representation has lower storage and processing cost, and retains the identification information of the original 3D face, thus it is helpful to the practical human recognition.