Hi,
I was looking at the Simba OCR today and I found out I pretty much forgot how it worked actually worked. (And I only wrote it a few months ago...)
Anyway, I went through the uptext code again and documented it. It probably has some typo's, let me know if you find any. It's not fully finished but the filtering is documented.
You can find it here: http://wizzup.org/simbadoc/mmlref/ocr.html
The full source to the OCR unit, it may help when trying to understand the documentation. http://git.villavu.com/?p=simba.git;...MLCore/ocr.pas
Fourth post in this thread has a picture attached of the OCRBench tool which displays how the OCR works. http://villavu.com/forum/showthread.php?t=53343
I'm not sure how old the picture actually is, but it should show how the filtering works anyway. If someone wants to play with the OCR tool, I can update it and host it somewhere.
Enjoy,
~Wizz