Theory behind the calibration of the depth camera intrinsics from rgbdemo

[Fernando Flórez Gómez]

Hi everyone, i am working now with libfreenect and rgbdemo  latest version, it run perfectly on Ubuntu 12.04, so checking the calibration I found the theory behind the RGB calibration, and I understand it very well, but when I try to look for information about how to Nicolas calibrate de depth camera I just don't get it well, here http://nicolas.burrus.name/index.php/Research/KinectCalibration  says that the main difficult comes to calibrate the depth cam cause it can't detect patterns on a flat surface, and it's obviously so he created the pattern using depth difference, and I wonder how do it? How we can fiund the intrinsics parameters like focal length, or distortion parameters using just that? He use zhang's technique for that too?  There is a theory behind this, where can I found it?  What is the relation between selecting the corners of the chessboard (like image below) on the depth image and the way he gets the intrinsic parameters?  I mean how the intrinsic parameters of the depth image are obtained, there is an article or link where I can find the theory?

Thanks for you opinion and I am sorry about my english, i really appreciate somebady could help me...

Thanks again! =)