Before you spend a lot of time on this, you might want to have a look at A little fun with merlyn and the discussion around it. In my opinion, it's just not all that hard to OCR even "obfuscated" text.

As far as munging images goes, GD seems to be a popular choice. (GD::SecurityImage seems to do what you want, in fact, though I don't know whether its images are sufficiently "obfuscated" for your taste. Is that the module you were talking about?) For procedural image generation, especially unique image generation, this book is the best reference I've ever found.

Going a bit off topic, I wonder what sort of "security images" one could come up with that would be reasonably secure, yet easily recognizable? You'd probably have to get into audience-specific content; for instance, a site for gamers might use screenshots from computer games, and ask the user to identify the level shown. Any ideas?