On Fri, 3 Jun 2005, m. allan noah wrote: > it says OCR is done by software. this means it is likely able to > take a traditional 'picture' of the document. this part could
But it only has to do OCR on the OCR-A font. This should be a simple task for which general-purpose OCR software and a 'traditional picture' would be overkill, and too slow. It reads the documents at half a metre per second, printing on them and stamping them as it goes. It would be a non-flatbed application in which the OCR characters are at exactly-known locations, and high throughput speed is possible. > ... all the other things it does (reading micr, etc) and the actual > OCR process cannot be done via sane, but perhaps by an external > program (which you would likely have to write yourself) In Cobol :-).
