As mentioned above, tesseract is an open source project that works very well. PDF::OCR, PDF::OCR2 are perl interfaces to it, makes some image conversion decisions for you, etc.
I've also tested out gnu ocrad, ran some benchmark and output tests on same material etc- it doesn't work as well.

This stuff works well on posix, these are mostly linux type boxes.

On windows there's something called iris - but it's pay, it costs a lot, and you have to pay extra for an sdk. So you can't really code with it out of the box.