3D Face Recognition

Abstract

Face recognition using 3D cameras captures and processes explicit 3D shape information and is more resistant to spoofing security attacks than 2D face recognition. However, typical compact-baseline, triangulation-based 3D capture systems are generally only suitable for close-range applications. The 3D shape may be used alone, or in combination with simultaneously captured, standard 2D color images in multimodal 2D/3D face recognition. Like its 2D counterpart, modern 3D face recognition often uses a feature extraction phase followed by a classification phase. Traditionally, such features have been handcrafted, but recent approaches use the representational learning power of deep neural networks. Other approaches use shape alignment methods, or analysis-by-synthesis, which exploits the prior shape distribution information captured by 3D morphable models. Additionally, 3D face models have been used for pose correction and calculation of the facial albedo map, which is independent of illumination. Finally, 3D face recognition has also achieved significant success toward expression invariance by modeling nonrigid surface deformations, removing facial expressions or by using parts-based face recognition. This chapter is not intended to be a comprehensive survey. Rather, it gives a tutorial introduction to 3D face recognition and a representative sample of both well-established and more recent state-of-the-art 3D face recognition techniques in terms of their implementation and expected performance on benchmark datasets.