its called homography... fairly strait forward to code...

http://en.wikipedia.org/wiki/Homography_(computer_vision)