Realistic 3D face modeling has become the focus of researches in computer graphics, but the existing researches have not provide a convenient, efficient, realistic 3D face models generation method. Based on the analysis of pose correction and transformation of 3D face model, a method of personalized 3D face modeling by matching a few key feature points is discussed, which need only a face picture and a common 3D face model. Then, a prototype system was implemented using C++ and OpenGL, and the experimental results suggest that the proposed method is easy and efficient, and can be widely used in film and advertising, 3D network games, and medical hairdressing, etc.