Individual 3D Face Generation Based on Candide-3 for Face Recognition

In this paper, an individual 3D face generation method has been proposed. This method is based on candide-3 and aims for pose normalization for face recognition system. By using ASM and AAM for facial feature point tracking and model parameters optimizing, this method decreases the computation time of 3D face generation and improves the level of the 3D model fitting with the input image. Through applying the method for pose normalization, the recognition rate can be improved greatly. Experiment results on CMU-PIE database show validity of this method.