HowToFixComputers.com

Frank Haber · Guest

OCR-ing always puzzles me. I still have Omnipage 11 here, and it's only
middling, with my 600/1200 scanners. Small type always flummoxes me and the
scanner.

In my ignorance, I've noticed several things.

1. Never scan in 2-bit. You get motheaten text that won't OCR.

2. Scan huge if the text is small. It might be worth even looking at the
histogram.

The absolute worst is type that's been photo-reduced previously. Think LP
record liner notes, reduced 2:1 to 7" square for a reel-reel tape or to 5" for
a CD. It's impossible. Even though the text is fully legible to the eye,
1200 is not enough. The results are laughable. Oddly, if I blow up the
original optically on an ancient copier, the scans then become only
semi-laughable. We're still at the 10% error level, though. It's easier to
retype.

I presume the pros licked this problem years ago. Would modern equipment and
software help?

And while we're at it, which of the current OCR programs do you think would
work best with content that's 40% English, 20% Spanish, 20% Italian, with the
rest usually in French or German (always in italic).