I've got a requirement to build a nutritional database of as many ingredients as I can get my hands on. Right now they have to be hand entered into a web app, but there has GOT to be a better way using OCR or some other technology I haven't thought of.
There are several OCR libraries on sourceforge, I've looked into and tested a couple, some don't work at all, while others work fairly well, but only from a PERFECT nutritional panel (photographs from a mobile phone don't work).
My question is, what would be the best approach for taking a picture of a nutritional label, and automatically gathering the information off of it? If it is photographed (mobile camera, or DSLR) is there anything that can be done to the picture to help make the OCR engines job easier?