Distortion correction for complex multi-projector systems.
The basic idea behind it is to project a series of black/white patterns at a screen, once for each available projector, and record those patterns with a camera. The patterns are chosen such that each pixel of a projector flashes a unique on/off code. This can be used to compute a regression model for the relationship between the position of a pixel in the projected image and the position on the screen. This software uses this regression model to generate binary mapping files which can be used in OpenGL programs to pre-distort an image such that it will appear undistorted to the camera, when projected.
This is currently not a complete program (or even great code), but rather a framework which may be filled in by the user or just used for inspiration.
We've used this code to implement the visual part of an audio-visual VR environment.
See http://www.tatome.de for a way to contact me.