Extracting PDF Figures as PDFs in Linux

The perennial problem: you have a PDF of a paper, and you want to extract one of its figures to show in a talk or something.

If you’re a normal person, you open it up big in your PDF viewer, take a screenshot, and call it a day. Or your PDF reader just lets you draw a box and save its contents.

But maybe you’re running Linux and don’t have commercial software available; and maybe you want to make a vector screenshot, not a bitmapped one, so that full detail will be maintained and you can scale the figure around. Evince doesn’t support draw-a-box-and-save-it, so you’re sunk.

But you’re not! You’ll need xpdf, poppler-utils, and pdfcrop — all widely-available. You can draw a box on the page in xpdf, and you can also set up xpdf to run a command when you press a key, with the box parameters able to be substituted into the command line. Then the pdftocairo utility provided by poppler-utils can rerender the page to PDF while cropping to your selected box. pdfcrop makes the margins nice.

You need to put this short helper script in your $PATH. It takes the coordinates from xpdf, converts them to the system needed by pdftocairo, then chooses an output filename (fig${n}.pdf) and makes the magic happen. There’s nothing complex about it and it can easily be monkeyed with to fit your preferences.

Then put this line in ~/.xpdfrc:

bind ctrl-e any "run(xpdf-extract-helper.sh '%f' %p %x %y %X %Y)"

Straightforward enough, I hope. I chose Control-E as the keybinding for “extract”.

Now just open up your PDFs in xpdf, draw boxes, and hit Control-E! Here’s a sample result … converted to a PNG so that it can render inline, so that in a way it misses the whole point of the exercise. But there’s little PDF figure behind it, I swear.

McLean et al (2012 ApJ 746 23), fig 2. I assert without proof that I have a PDF version of this image.