3D Face Reconstruction for Single Visual Images

This paper proposes a fast and efficient method of 3D face reconstruction for single visual images which are obtained under uncontrolled conditions.First of all,face feature points are located through the ASMS(active shape model of subset).During this locating processing,ASMS locating accuracy is improved by increasing the completeness and the clustering performance of the training samples,which also sets up the fundamental basis for further face feature extraction.Secondly,2D-3D feature subspaces are built up according to different subsets respectively using method LFA(local feature analysis),which efficiently fuses the global and local features.Thirdly,a dynamic selection strategy for 2D-3D features subspace selection is designed,which can significantly improve the reconstruction accuracy and keep a content processing efficiency at the same time.What is more,Experimental results show that the proposed method is not only simple and efficient,but also robust to pose and illumination variations,and achieves desired reconstruction results on WHU-3DFaceData database.