Well, the advantage that screen scraping has over paper OCR is that the font is completely deterministic and there's (usually) no degrading effects. Even if '5'

Well, the advantage that screen scraping has over paper OCR is that the font is completely deterministic and there's (usually) no degrading effects. Even if '5' and '6' are pretty similar, there must be some pixel that can be used to distinguish them. What could make it more tricky is if alpha transparency is used for both the foreground and background. Still, the complexity isn't anywhere near "real life" OCR.