SIFT implementation of OCR in MATLAB

Question

So for our Final Year Project we are making a nifty lil app that lets a user take pictures of visiting cards from their cell phone, apply OCR on it and get all the required info in printable form... My problem is that SIFT is rotation invariant i.e it can not differentiate between a "u" and an "n" and other such words and numbers. I would really appreciate if you share any ideas that you folks mite have or come up with for overcoming this problem. :) Thank you.

Answer 1

Try to strip that image then use dilute, if that does not work use neural network

Answer 2

Have you looked at the following paper? :-)
Scale and rotation invariant recognition of cursive Pashto script using SIFT features

Answer 3

It is true that SIFT will have difficulties working out 'u' and 'n' similar characters but if you really want to stick with SIFT (your final project is probably finished now...) you could always apply some pre-processing on images like 'u' and 'n' by putting underlines under them in right orientation and let SIFT know that 'u' with underline is 'u' etc.