Here is a suggestion for a different setup.
I tried to build an MT Mini by your instructions, but got quite dissatisfied with the result. First of all, the manual goes mainly for Windows. I figured how to build touchlib for Linux, but the final running program wasn’t showing the same windows and kind of missing stuff. I fired up Windows and did everything there. Although the table worked, I immediately noticed two shortcomings:
a.) Illumination requirements are picky. It works only best if the light is scattered across the room so the shadows become blurry as soon as the finger is not directly touching the surface. I have halogen lamps in my room, so that’s a real problem. I tried covering them a bit, but then the light becomes too dark for the camera to see enough.
b.) If finger shadows overlap, there is a problem. This is also part of a.). If the light comes more from the side, some finger shadows overlap, you have to constantly watch how you spread your hand and move your fingers. Not every constellation works.
Additionally I have to say that controlling the desktop is also problematic, since the surface I am controlling is somewhere else than the surface that is being controlled. Of course this one is obvious.
So it got me thinking how to improve the setup and fix some of the issues. I think I found a much better solution that requires also only a Webcam, two cardboard frames, some LEDs and excessive amounts of tape, although I wanted to validate this idea with you before I seriously start making a plan.
1. Cut out two cardboard frames so that they match the size of your computer screen. The frames borders should be wide.
2. Like with the big MT table, place LEDs between the frame borders at the top and at the bottom, away from the inward border, so that when you place a piece of paper inside the frame, the light cast by the LEDs casts a hard shadow which seems to divide front and backside of the frame.
3. Place the frame so that it hovers in front of the screen, at a distance of say 10cm, although more is usually better.
4. Place the webcam high above and a little bit behind the screen so it looks at the back of the frame.
Now when you put your fingers through the frame, your fingertips will be illuminated - this corresponds to a touch. The camera can clearly distinguish an illuminated fingertip from the background. The image has to be perspective corrected before being processed, of course, and the vertical resolution will thus be a bit lower than the horizontal one.
The result is that you can turn your screen into a touch screen this way - the less distance the frame has to the screen (which requires more perspective correction and lowers the vertical resolution even more), the more it becomes a touch screen.
I think you could also use infrared light here - even a glass window, like in the big table.
It’s just speculation but I think I made no mistake. Please let me know what you think