Automatic 3D facial model reconstruction from single front-view image
Yanchao Xing, Chunxia Guo, Zheng Tan · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2004
A novel algorithm for automatic 3D facial model reconstruction form a single front-view facial image is presented. First, face and facial features are located. Then feature point depths are estimated from frontal facial measurements through multi-variant linear regression. With these feature points, the generic face model is individualized by volume morphing with a novel Radial Basis Function-based interpolation. Finally, the texture map is generated through cylindrical mapping. Experiments show that the reconstructed model has satisfying visual effects for a broad pose range. Compared with the multi-view based facial model reconstruction, this algorithm is fast, easy to use, fully automatic and has low requirement in hardware, so it has great potential for many practical applications.