Qifa Ke and Takeo Kanade December 2006
In Proceedings IEEE Conference on Computer Vision and Pattern Recognition (CVPR)