Should be fairly easy... But normal OCRs are designed to recognize printed text, not VGA font, so I'm not surprised they fail. However, if we know the font used, recognition should be simple. (Bad debian and their non-standard font...). Everything could be probably helped by having something like "XXXXXXXX" line at the top of oops so that decoder could calibrate itself. Or maybe existing parentheses around addresses could be used. Would it be enough to decode just the numbers?