[Ridcully]

I have a DJVU file that I want to read on Kindle, which has high quality scans with little noise, all properly oriented etc. However, when I try to convert it to any other format (including PDF) with Calibre, I end up with the underlying OCR text, which is horrible - no paragraph breaks, probably never checked for spelling, any bold/italic text is completely garbled... How do I get rid of this and get the scanned images instead?