GoCam is an open source (GPL) project for developing a computer vision system that analyses video capture (for example with a webcam) of a go game and extracts the moves made by the players. It is implemented in C++ and uses the excellent CImg Library for image processing. Currently, the system is able to find the playing grid from still images.
For more info visit its website at http://iki.fi/thirsima/gocam/. (dead link)
Here you have an example:
The latest release (0.3) does not build on a modern system. A git repository to collect any patches are available at http://repo.or.cz/w/gocam.git/. This slightly updated gocam segfaults on several input images, and misidentifies the grid for many. I am not sure if this is due to using newer tools and libs, but there were already less segfaults and wrong results when switching CImg 1.2 (gocam 0.3 ships an old version of CImg).