Found Some Information ,
Also Here :
The letter, or word, is compared to a selection of letters, and compare the different and relative factors of shape, black vs white, and holes. It then develops a set of rules and then a system of comparison. That would be ICR, but OCR is more based on a set example.
CAPTCHA uses words from books that OCR has a hard time to read, morphs and adds patterns to the letter, to further obscure the ability for computer readers.
It's effective, but if not many set word CAPTCHAs are made, it can be manually solved by a person that is trying to bot it, program the bot that the CAPTCHA means so and so, and the CAPTCHA is useless. There are also people paid in developing countries to solve CAPTCHAs.
Sorry if I got off-topic and/or useless. ;_;