Qifa Ke and Takeo Kanade 2004
In Proceedings IEEE Conference on Computer Vision and Pattern Recognition (CVPR)