Yes it's entirely possible.
One of the first things to do with cracking some of these things is to remove the color s.t. the image is full black-and-white. Then, some set of steps is followed to determine the text content of the image.

The original captchas that google used widely before implementing the reCaptcha system are considered easily crackable and not a very good detirrent for a determined cracker (and a lot of software which helps out uses *machine-learning* and *neural* *networks* and lots more to determine the text content of an image)

desu I really miss the old captcha system becuase of shit like: http://i.imgur.com/oCa4d.png

>>6827
machine-learning AND neural networks you say?

>>6828
OCR uses coding and algorithms to detect text content

dragon ball z