3D Face Reconstruction Based On One Image

3D face reconstruction is a hot research topic in computer animation, computer games, computer vision, and image recognition fields in recent years. It is also the research focus in computer graphics and computer vision, and has many challenging open problems. The traditional 3D face modeling methods need several face images, and then modify a general face model. Modeling 3D face from one image is a new way which was proposed several years ago. This paper concentrated on the 3D face reconstruction from one image.Firstly, the 3D model data was got from the software and the face images of CVL Face Database, and the data was standardized to adapt to the study.Secondly, the paper introduced the method of 3D face reconstruction from one image. The face pose parameter was obtained based on the face pose estimated algorithm. Based on PCA and Sparse Morphable Model, the 3D shpe model was reconstructed, and the RBF interpolation method was used to ensure that the model was more precise.Finally, the correspondence dense 3D face was achieved by mesh subdivision, and the texture parameter was optimized by the modified BFGS algorithm based on the new quasi-newton equation to reconstruct the face texture.The 3D face reconstruction was implemented ultimately.This paper carried out simulation tests using Matlab in the corresponding chapters, solved some key techniques in face modeling, and implemented the 3D face reconstruction from one image. At last, we analysised the existing problems, and provided research directions and some experiences for the further study.