基于单张照片的三维人脸重建优化算法

摘要

传统人脸三维重建算法难以确定人脸形状,并且计算复杂.针对此问题,提出一种以水平集方法获取人脸轮廓并结合明暗恢复形状(SFS)算法重建三维模型的方法,该方法仅需单张正面人脸照片.首先采用主动形状模型确定人脸轮廓,将其作为水平集的初始演化曲线,分割出完整的人脸形状；然后对人脸区域进行灰度变换,求出灰度图像；最后通过SFS算法重建已知光照条件的人脸图像的三维模型,将该模型作为参考与灰度图像匹配,进而确定其光照条件和三维模型.实验结果表明,与基于网格模型的算法相比,该方法可快速地重建具有完整形状的人脸模型.%The original 3D facial reconstruction algorithm cannot determine the shape of face,and it is also very complex.To resolve this problem,a 3D facial reconstruction algorithm based on the Shape From Shading(SFS) by a single photo was proposed,which used the level set method to determine the facial contour.Firsdy,initial contour of the face was defined by Active Shape Model(ASM),and it was used as the input of the level set method to segment full face shape.Secondly,the facial image was converted to grayscale.Finally,a reference model was established by using SFS on an image whose illumination condition was known.The input image was matched with the reference model,and then the illumination condition of the input image and the reconstruction result could be calculated.The experimental results show that,compared with the mesh model-based algorithm,the proposed method can ensure the shape of face and obtain the reconstruction model in a short time.