@darius the best bets with Topaz are either extracting the low-quality OCR text, or converting it to an image-based PDF. I think tools exist for the former, haven't seen anything for the latter.