Yesterday I registered on a forum and at the bottom of the registration appeared something that I´ve seen several times lately, a little image with a word in it which I have to write on a text box.
Now, I understand that this is probably to prevent automatic registering (to avoid mass spamming for example), so a program wouldn´t be able to read a word on an imgage; so far, so good. What I don´t quite understand is why the word and image appear distorted/scrambled in many ways, I´ve see several types of distortion used; why is that?
Exactly. Humans are very good at recognizing distorted shapes and computers are not. If they don’t distort the letters (enough), OCR programs are able recognize them pretty safely.
Your WAG on OCR is correct. It is quite easy to OCR a text image that is just black on white using a sleek font. Ugly up the background, jiggle the letters. Makes it much harder.
Note that many “fringe” sites now require people to decode such images for them for the purpose of aiding spammers! They want a Yahoo! account. Yahoo! feeds them the text image. They put the text image on a registration page. The next person to sign up gives them the answer. They feed that back to Yahoo! and get an account to spam with.
Unless there definitely appeared to be legitimate reason for them to require decoding such an image, I would not aid possible scumbags.
Another factor is that for this use, your OCR (“optical character recognition”) doesn’t even have to be that accurate. If you get 1 out of 10 words right, that’s still probably good enough for the spammers. So the words have to be distorted almost to the point of illegibility.