I have experience to build a document management system in which i was using open source OCR to extract characters from image file,. But i am sorry to say you that 100% accuracy of OCR is not possible according to current technology. If you accept that, then i will be happy to help you to complete your stuff.