For a scanned book with the two sides together, I would use either
PDFLRF and/or
PaperCrop.
The best results I've gotten have come from using cropping out the individual columns with Acrobat as separate pages, and then going through PDFLRF, but that's a lot more time consuming, and not feasible if you've got dozens of hundreds-pages-long books.
I believe
PDFRead, which is another possibility, has an option to call unpaper while processing, but personally I've gotten better results with PDFLRF than with PDFRead.