Automatic 3D Face Modeling with Single Image and Perspective Projection Model

In this paper, a precise and robust algorithm is presented to automatically model the 3D face with only single frontal face image. For shape and projection parameters estimation, an energy function is built to represent the difference between the approximate observed face 3D shape and estimated face 3D shape. To achieve more robust results, shape constraint is added to the energy function. For precision and reality, the perspective projection model is adopted in both 3D shape and 2D texture estimation. The experimental results show that the 3D face modeling is very robust and precise. And it could be applied to the most popular desktop camera systems and mobile-phone camera systems.