I’ve solved this problem in C#. Wrote my own little library (to keep the fingers in shape). Functionally, think about what you would do manually to scan a picture and compare it to another. One part of my library learns what the characters in the images look like, the other part uses that information to recognize other characters. I was able to use this library to solve the first 3 captcha levels. Almost done with Captcha 4, just one nasty glitch in the way standard .NET processes images.
For me this was more of a coding-practice than the Coding levels :) As always (not just with coding): Think what you want to achieve, think about how to achieve it functionally, then think about a way to let the computer handle that functionality and test it before you use it.
Good luck!
