Normalized 3D to 2D model-based facial image synthesis for 2D model-based face recognition

In our previous research, we created a database of 3D textured face models of people using stereo images and a generic face mesh model for 3D face recognition application. Consequently, in this paper we make use of this available database and propose an algorithm for synthesizing multiple view 2D facial images of each subject, which extends the number of images used in the training stage of a 2D face recognition system. The main contributions of our work are: a) proposing a novel 3D model-based face pose and scale normalizations before creating the synthesized 2D database from its 3D counterpart and b) proposing a model-based facial area segmentation and normalization to a given 2D probe facial image. Recognition experiments, using near frontal probe facial images and the extended synthesized database, demonstrate improved 2D recognition rate.