I want to know if there are some methods to recognize letters (my case is a little easier, I need to recognize only capital letters) without using an OCR.
At the moment my application can produce a binary version of each character, not very precise due to the bad resolution of the image, but good for the most of the images. So I need a computational fast and reliable algorithm to recognize it.
How can I do it?
try to compare a Hu moments of charecter images http://www.academia.edu/289447/Multi-Font_English_Character_Recognition_Based_on_Modified_Invariant_Moments