I'll explain more... later. For the time being, I wanted to implement a sparse matching program, looking for at a limited number of potential matching points. To make it faster, there are two choices... use fewer points or use more cogs.
This is not an original approach or my favorite matching algorithm (waiting for SDRAM to start working on that). But it is fast and good enough to find a number of good matches that far outweighs the number of false matches.
This approach will eventually be useful for stereo-ranging in real time. As it is, there are too many false matches (not very many... just too many).
I have experimented with changing the parameters around and there seems to be no combination that eliminates the noise completely.
The next step is to elaborate the logic a little tiny bit. I think that will do the trick.
To see it, set the baud rate in P2host1x.spin to whatever your terminal supports... then run it and hit the zero (0) key.