From Few to Many: 
Generative Models for Recognition Under Variable Pose and Illumination

Athinodoros S. Georghiades, Peter N. Belhumeur, David J. Kriegman

abstract
We present recent work on appearance-based methods for modeling  the variability in the images of objects due to pose and illumination.  Unlike previous appearance-based approaches, the presented method requires  only a small set of training images to generate a rich representation that  models this variability. From as few as three images of an object in fixed  pose with varying but unknown lighting, a surface and an albedo map are  reconstructed and then synthetic images are rendered under variable pose  and lighting. A representation useful for recognition is thus generated. The methods have been tested within the domain of face recognition on  a database of 4050 images of 10 faces with variable pose and lighting; their performance exceeds that of popular existing methods.