3D

Abstract

Much of the existing face recognition systems operate in 2D. As the search for improved recognition performance intensifies, alternative methods, e.g., 3D, are under consideration. Head and face modeling using 3D could also benefit novel applications related to animation, digital film technology, game development, and medicine. The perceived advantage for 3D comes from its ability to provide depth information, and at the same time yield a better grasp on the overall shape of the head the face belongs to. One could also seek and chart manifolds in 3D that capture the intrinsic characteristics of the spatial-temporal trajectories traced together by the head and face. Another perceived advantage comes from the perception that “working in 3D overcomes limitations due to viewpoint and illumination” (Medioni and Waupotitsch, 2003). More is not always better if attention is not paid to the new imaging medium regarding enrollment and variability. 3D data acquisition, while not dependent on surrounding illumination depends, however, on the illumination source used for image capture. 3D enrollment considers [easy] availability [of imaging devices for capture purposes], capture and costs, on one side, and consistency, on the other side. Consistency concerns repeatability of the measurements obtained. Depth varies, however, according to facial expression, head size, stress, diet, physiological state, and aging. The face is quite complex both in appearance and the way it changes in 3D. The spatial and temporal challenges still persist and have yet to be met. The context for both challenges needs to include large scale data sets, in the thousands, rather than small and limited galleries. The gallery's composition should also display diversity. The two challenges could possibly be addressed together using spatial-temporal analysis (see Sect. 7.8).