There have been several discussions on OCR (Optical Character Recognition) in the Monastery that I can think of. You might find some of them helpful.

jcwren wrote a little snippet that he posted here which performs very basic OCR on an image file. It's quite limited in its capability, but demonstrates some of the principles involved.

The gyst of other threads on the subject seems to be that OCR is a computational-intensive task, and probably better suited to a compiled language.

The fact that there are no modules on CPAN for OCR (at least none that my quick search turned up) would seem to indicate that it's no trivial task.

Update: There is an Image Processing Algorithms module on CPAN that you'll find here which proports to be "designed for solving image analysis and object recognition tasks in Perl." The Quantum Leap of developing an OCR project from scratch may be reduced to an Enormous Hurdle with the aid of this IPA module.