One of the things that tells humans and computers apart /is/ distinguishing characters. Basic
vision is one of the issues that drives modern AI research. It was several years old, but one
test of OCR programs at UNLV showed that the best OCR program was inferior to a six year old
in correctly recognizing text. Currently in transcribing etexts, Distributed Proofreaders
(which transcribes for Project Gutenberg) does a opening level of OCR, but there's still a lot
of cleanup for random volunteers to do.