单幅图像融合参考模型的三维人脸重建

基于单幅图像的三维人脸重建以其数据获取简易而具有广阔的应用前景.本文提出了一种融合参考模型的方法实现基于单幅图像的三维人脸重建,其主要思想是以球面调和函数表示图像的光照模型,从而建立图像的光照信息与深度信息之间的解析关系;进而可基于参考模型的深度信息获得输入图像的近似光照系数,再根据近似所得的光照系数恢复其深度信息.该法不仅可完全自动化地实现,而且因不需要多个人脸模型的信息而优于基于统计模型的三维重建;由实验中对不同库中人脸图像及真实采集的人脸图像进行重建的结果发现,本文提出的重建方法是切实可行的.

3D face reconstruction based on a single image with reference model

There are broad application prospects for 3D face reconstruction using a single image as it isconvenient to capture. This paper employs a 3D reference model fused with a single image for 3D facereconstruction. ;he key idea is to connect the image lighting information with depth information byspherical harmonic function. ;hus the lighting coefficients can be approximated through the depth in-formation of reference model. Then the depth information of input image is captured based on the ap-proximated lighting coefficients. The proposed method can not only get rid of the flaw of shape fromshading method, but also do not need numbers of facial models which are used in statistic models.Therefore, the proposed strategy is convenient and effective for applications when the results are basedon the images in Bosphorus database,JAFFE database as well as the data captured in real world.