Geometric invariants for 2D/3D face recognition

In the last decade, security aspects such as biometrics have become one of the most central topics for governments as well as researchers, while the availability of more and more advanced technologies at lower costs has made image and video analysis also applicable for this aim. In particular, the 2D image analysis has been widely used in trying to overcome the main drawbacks of the face biometric (pose and illumination). Face is more attractive than most other biometrics, since it is fairly easy to use and well accepted by people, even if not yet robust enough to be used in most practical security applications. One possible way of overcoming this limitation is to work in 3D instead of 2D. But 3D is costly and more difficult to manipulate and then ineffective in authenticating people in most contexts. Hence, to solve this problem, a novel face recognition approach is proposed, using an asymmetric protocol: enrollment in 3D but identification performed from 2D images. So that the goal is to make more robust face recognition while keeping the system practical. To make this 3D/2D approach possible, geometric invariants used in computer vision are introduced within the context of face recognition. Results obtained in terms of identification rate are encouraging.