3D Face Reconstruction from Single 2D Image Based on Robust Facial Feature Points Extraction and Generic Wire Frame Model

An efficient 3D face reconstruction method for any single 2D facial image even partially damaged or hidden is proposed in this paper. Our purpose is to overcome the shortcomings of traditional methods such as providing at least couple of orthogonal facial images, needing complicated calculations, and so on. Especially, most of the existing methods can’t be used in real-time communication and mobile computing environment, e.g. video transmission systems. Firstly, some facial feature points were extracted by an improved active appearance models, we do some significant work on robust feature point localization when the facial images are not very well for implementation of the AAMs algorithm. Then we adjusted overall and local generic wire frame model using the feature points. Finally we constructed realistic facial texture. Experimental results show that our proposed 3D reconstruction method cost little time to model from a single image and is suitable for real-time and mobile applications.