Qifa Ke and Takeo Kanade December 2001
In Proceedings IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
