Skip to Main Content
We introduce a novel technique for face recognition by using 4D face data that has been reconstructed from a stereo camera system. The 4D face data consists of a dense 3D mesh of vertices describing the facial geometry as well as a 2D texture map describing the facial appearance of each subject. The combination of geometry and texture information produces a complete photo-realistic model of each face. We propose a recognition algorithm based on two steps: The first step involves a 3D or 4D rigid registration of the faces. In the second step we introduce and evaluate different similarity metrics that measure the distance between pairs of closest points on two faces. A key advantage of the proposed technique is the fact that it can capture facial variations irrespective of the posture of the subject. We use this technique on 3D surface and texture data comprising 62 subjects at various postures and emotional expressions. Our results demonstrate that for subjects that look straight into the camera the recognition rate significantly increases when texture and geometry are combined in a 4D similarity metric.