I'm curious about the algorithm and basic principles behind Optical Character Recognition.

Can anyone explain:

  • The most basic OCR algorithm that can be fairly easily implemented from scratch? Or the common denominator between different OCR algorithms?
  • Extensions/changes made to tailor the algorithm for application specific goals?

Also see my follow up question on OCR.

  • 3
    $\begingroup$ This question is a little too broad. You're practically asking for a book reference. Narrowing it down a bit would help getting answers. For what it's worth, do some research on number 1 (hint: the answer is that there are many different algorithms with different steps). The second bullet point is perhaps the best basis for a question. The third could be a follow-up question all by itself. $\endgroup$ – Phonon Jul 31 '12 at 15:46
  • $\begingroup$ The simplest scheme is probably the original MICR scheme, where a special font is used that is designed to be easy to read, and where the lines of characters are cleanly aligned under the reader with BOT/EOT markers. $\endgroup$ – Daniel R Hicks Aug 1 '12 at 0:07
  • $\begingroup$ @Phonon, I created a follow up question, and changed the question a bit to narrow the scope. Hope this is better now. $\endgroup$ – Geerten Aug 1 '12 at 6:55

Probably the simplest OCR algorithm would be the case where the characters are from a known font, in perfect rows with no distortion or rotation, and then you can find occurrences of each character in the text by doing cross-correlation with the known characters from the known font.

Or even simpler if you can create the font before it's used, with cheaper ways to recognize characters, like the MICR fonts or OCR fonts.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.