Click to See Complete Forum and Search --> : internet bots and images


abcdefgqwerty
March 20th, 2007, 10:00 PM
No one has yet invented a way for internet bots to read the numbers out of textboxes that contain the same colors as the distorted numbers and letters? Thats hard to believe it seems possible.

Zaccheus
March 21st, 2007, 04:20 AM
It's an arms race: A company invents a way of obscuring the letters, another company finds ways of reading the obscured letters, so a thrid company invents new ways of obscuring the letters, etc. Also the bot needs to somehow know which image contains obscured letters and which text box the code needs to be written into.

GNiewerth
March 21st, 2007, 05:50 AM
Hi,

it might be a question of processor time, too.
It´s takes almost no time to get the text from a textbox, but you´ll need much more time to analyze an image and perform OCR. This might be neglectible for a few operations per second, but can be a serious bottleneck when handling a few thousand images.

Regards,
Guido

rdrast
March 21st, 2007, 07:44 AM
http://en.wikipedia.org/wiki/Captcha

http://www.cs.sfu.ca/~mori/research/gimpy/