A fundamental problem in multi-view 3D face modeling is the determination of the set of optimal views (poses) required for accurate 3D shape estimation of a generic face. There is no analytical solution to this problem, instead (partial) solutions require (near) exhaustive combinatorial search, hence the inherent computational difficulty of this task. We build on our previous modeling framework which uses an efficient contour-based silhouette matching method and extend it by aggressive pruning of the view-sphere with view clustering and imposing various imaging constraints. A multiview optimization search is performed using both model-based (eigenhead) and datadriven (visual hull) methods, yielding comparable best views. These constitute the first reported set of optimal views for 3D face shape capture and should provide useful empirical guidelines for the design of 3D face recognition systems.
Published in: IEEE Int'l Conf. on Automatic Face & Gesutre Recognition (FG'04)This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved.