In 2007 through 2009 we carried out a series of studies of factors that influenced the performance of three algorithms from the Face Recogntion Vendor Test 2006. The results of this study are summarized in the following papers.
Focus on Quality, Predicting FRVT 2006 Performance, J. Ross Beveridge, Geof H. Givens, P. Jonathon Phillips, Bruce A. Draper and Yui Man Lui, 2008 8th IEEE International Conference on Automatic Face and Gesture Recognition, September 17--19 2008, Amsterdam, The Netherlands. (Received the Best Paper Award)
FRVT 2006: Quo Vadis Face Quality, J. Ross Beveridge, Geof H. Givens, P. Jonathon Phillips, Bruce A. Draper, David S. Bolme and Yui Man Lui, To appear in Image and Vision Computing ( doi:10.1016/j.imavis.2009.09.005).
To highlight just one finding from these studies, edge density in an image is, in many cases, an extremely strong predictor of face recognition performance. For example, when comparing a new image taken outdoors to a stored image taken indoors under controlled conditions, the higher the edge density in the new image the less likely it is that a face recognition algorithm will correctly recognize ther person.
