Dr. P. Jonathon Phillips, National Institute of Standards & Technology
Face recognition is an interdisciplinary field with elements from image processing, pattern recognition, computer vision, computer graphics, psychology, and evaluation methods. All these areas have influenced the development of automatic face recognition algorithms. My survey talk will discuss how each of these areas has influenced the development of face recognition algorithms.
Interest in face recognition algorithms exploded in the 1990s. Significant approaches developed during that timeframe were subspace techniques and elastic bunch graph matching. Examples of subspace techniques are algorithms based on principal component analysis (eigenfaces) and fisher discriminant analysis. More recently, under the Face Recognition Grand Challenge, kernel methods proved to be successful for recognizing faces across changes in illumination. Manifold techniques are being investigated to increase robustness to recognition across illumination and pose.
Three-dimensional morphable face models use computer graphics techniques to model changes in pose and illumination. This technique has been successfully applied to improve face recognition across different poses.
Humans are extremely robust at recognizing familiar faces across changes in illumination and pose. However, this robustness does not extend to recognizing unfamiliar faces. Recent results show that machines are capable of outperforming humans on recognizing unfamiliar faces across illumination changes. In addition, machine algorithms have helped to explain properties of the human face recognition system and humans have provided input into improving machine algorithms.
Since 1993, the Face Recognition Technology (FERET) and the Face Recognition Vendor Test (FRVT) series of evaluations have measured progress in face recognition. These evaluations document the improvement of face recognition algorithms over the last 14 years and assess the state-of-the-art today. On frontal face images, performance has improved from a false reject rate of 79% at a false accept rate of 0.1% in 1993, to a false reject rate of 1% at a false accept rate of 0.1% in 2006. This is roughly two orders of magnitude decrease in the error rate.
Dr. Jonathon Phillips is a leading technologist in the fields of computer vision, biometrics, face recognition, and human identification. He is at National Institute of Standards and Technology (NIST), where is he program manager for the Face Recognition Grand Challenge and Iris Challenge Evaluation (ICE), and test director for the Face Recognition Vendor Test (FRVT) 2006. From 2000-2004, Dr. Phillips was assigned to the Defense Advanced Projects Agency (DARPA) as program manager for the Human Identification at a Distance program. He was test director for the FRVT 2002. For his work on FRVT 2002 he was awarded the Dept. of Commerce Gold Medal. His current research interests include computer vision, face recognition, biometrics, and computational psycho-physics. His work has been reported in print media of record including the New York Times and the Economist. Prior to joining NIST, he was at the US Army Research Laboratory. He received his Ph.D. in operations research from Rutgers University. Dr. Phillips is an Associate Editor for IEEE Trans. on Pattern Analysis and Machine Intelligence. He was co-guest editor of the Proceedings of the IEEE special issue on biometrics. He is a Senior Member of the IEEE.