Played with RPi Camera Board recently to see if it can be used for simple image-recognition.

Here is the result:

This is produced by this python script using picamera module to capture the image, ZBar bar code reader to find qr-codes and pygame was used to draw stuff. Latency is around 1 sec using 1024x768 image size.

I've realized there exists a problem with using see-through augmented-reality displays: if you need to draw something over a real object, you need to account for the camera offset from the display, which is important for close distances. I don't know how to calculate the correction dynamically without some kind of distance-detection hardware (the best would be to use a depth-sense camera... maybe they need to make one which is RPi-compatible).

I'll try to make a small demo video soon.