Scan Tailor is a great program for working with scanned images (of books). It deals with the common problems one has with scans: they may be tilted, they may be warped (e.g. curve towards the middle of the book), etc.

Sometimes, a PDF is made up of scanned images. (As opposed to being made up of characters laid out on the page.) In such cases, it is an attractive idea to “disassemble” the PDF into images, process them with ScanTailor, and re-assemble into a better PDF (or some other format).

Installing ScanTailor

The ScanTailor project has not seen updates in the last few years, so this was quite hard,1 at least on macOS. I later came to know (from some posts on diybookscanner.org) that there are forks of ScanTailor which have more features and/or may have better updates, but I didn’t know it at the time. I used these instructions to install (not a direct quote):

First, run

git clone https://github.com/Homebrew/homebrew-boneyard
cd homebrew-boneyard
git checkout f0f4e12197

Then edit homebrew-boneyard/scantailor.rb to change the line depends_on "qt" to depends_on "cartr/qt4/qt@4". Finally run:

brew tap cartr/qt4
brew tap-pin cartr/qt4
brew install qt@4
brew install scantailor.rb
scantailor

(Actually those instructions didn’t quite work for me directly, but I was able to figure out from there…)

Extracting images from PDF

From rendering

There are many ways to do it. For example (from here) one can “render” each PDF page, and write out the pixels:

gs -dNOPAUSE -sDEVICE=jpeg -r144 -sOutputFile=p%03d.jpg file.pdf

or

gs -dNOPAUSE -sDEVICE=pngalpha -r1260 -sOutputFile=p%03d.jpg myPDFfile.pdf

This will work regardless of whether the PDF is made of scanned images or not. The problem is that this “resamples” — e.g. if the PDF contains images at a certain resolution, the output of this is the outcome of rendering the page at (probably) a different resolution.

Directly

If the PDF really contains images, then we should be able to get the images directly out of the PDF stream. (I vaguely remember once doing it by just opening the PDF in Emacs and copy-pasting the relevant binary sections into new files…)

There are some relevant questions on Stack Overflow: See Extract images from PDF without resampling, in python? and linked questions.

(There are also questions about PDF structure in general: see inspecting PDF files and Convert PDF to a tree.)

Anyway, I ended up using pdfimages from the Poppler library. For example:

pdfimages -f 1 -l 230 -png myPDFfile.pdf someFooPrefix

or

pdfimages -all -f 1 -l 84 myPDFfile.pdf someFooPrefix

Added later: It appears one can also use mutool extract. Will try next time.

Using ScanTailor

Once you have the images extracted in whatever way, you can use either the ScanTailor GUI or scantailor-cli (I’ve had better results with the GUI, because there are things you don’t realize until you see the output).

Assembling images

I want pages laid out 2x2 on each page (to save paper), for which I can use, for example, montage from ImageMagick:

montage extr-007.tif extr-008.tif extr-009.tif extr-010.tif -geometry +1+1 montage.png

For example (the multiple variable assignment in the for loop requires Zsh I think):

for a b c d in {005..088}; do echo "$a and $b and $c and $d"; montage extr-$a.tif extr-$b.tif extr-$c.tif extr-$d.tif -geometry +1+1 montage-$a.png; done

Converting to PDF

ImageMagick can do that too:

convert montage-0* together.pdf

That’s it!

  1. It is unfortunate that we live in a world where already-working software needs to keep getting updated, just to keep working. It’s because most software has many dependencies (other libraries, at least the operating system), and those keep getting updated. That’s one reason it’s good to minimize dependencies. ↩