The main papers are Face detection and recognition using maximum likelihood classifiers on Gabor graphs and Face recognition by elastic bunch graph matching. They describe how Gabor Wavelets can be used to specify arbitrary textures; especially faces and how to recognize them. The technique is suprisingly easy to implement, very robust and outperforms lots of other systems on the market, especially on different lighting conditions, which is a problem on its own. I used a stripped down version in my BSc-thesis (C++/OpenCV) for head pose recognition; if you read pp.16-23, you get a basic idea about it.