The picture to text to speach part is definitely doable - you can even see something similar to what you described in action if you have access to a phone with Android on it - check out Google Goggles here:
Click[
^]
As far as I know Goggles dosn't actually read the OCR'd text out loud fr you, but I'm sure there's an app for that ;)
I would however like to point out that a blind person can't see what the camera is exactly pointed at... So you might get partial texts, crooked letters, all sorts of strange things to cope with. It might be quite complicated to use in the scenario you described. And it may take more than a couple of tries for the user to get the desired result - I think it might be a bit frustrating frankly... But hey - there are more uses for software like that. You can always use it as a language improving tool for students or something.
Whether or not it's a good graduation project is a different question. If you have the time and the skills - sure, I would give it the :thumbsup: as a very interesting idea :) Good luck whatever you decide!