I just had to prepare a poster based on roughly 30 publications, and for several of them I didn't have the original figures but only the manuscript as a pdf file. Using okular and a magnification of 800%, I've got screenshots of these figures as comparatively highly resolved bitmaps, but the price I had to pay was that the editing of the poster in LibreOffice (which I've used as the least common denominator) was getting almost unbearably slow. I couldn't silence the thought that it should be possible to take a 'vector screenshot' from a pdf file. I had the vague idea that pdftocairo could be useful in this respect, since it can output arbitrary parts of a pdf file as pdf or svg. And it turned out that [Peter Williams,](http://newton.cx/~peter/about-me/ "Peter Williams,") a young radio astronomer from Harvard, had the same idea and came up with a [script](http://newton.cx/~peter/2012/10/extracting-pdf-figures-as-pdfs-in-linux/ "script") which does exactly what I wanted. I've fixed a small error (_pageh_ should also be an integer) and ensured Arch and Fedora compatibility (python2), but otherwise its Peter's script: #! /bin/bash # original: # see margin=1 # XPDF gives its y coordinates in terms of the standard PDF coordinate # system, where (0,0) is the bottom left corner and y increases going # up. But pdftocairo uses Cairo coordinates, in which (0,0) is the top # left corner and y increases going down. We can use pdfinfo to get # the page size to translate between these conventions. file="$1" page="$2" pageh=$(pdfinfo -f $page -l $page "$file" |grep '^Page.*size' \ |sed -e 's/.* x ' -e 's/pts.*$') # Our variables end up in Cairo convention, so the box height is ybr - # ytl. xtl=$(python2 -c "import math; print int (math.floor ($3))") ytl=$(python2 -c "import math; print int ($pageh) - int (math.ceil ($4))") xbr=$(python2 -c "import math; print int (math.ceil ($5))") ybr=$(python2 -c "import math; print int ($pageh) - int (math.floor ($6))") w=$(python2 -c "print $xbr - $xtl") h=$(python2 -c "print $ybr - $ytl") # Lamebrained uniqifying of output filename. n=1 while [ -f fig$n.pdf ] ; do n=$((n + 1)) done # OK to go. echo pdftocairo -pdf -f $page -l $page -x $xtl -y $ytl -W $w -H $h \ -paperw $w -paperh $h "$file" '|' pdfcrop --margin $margin fig$n.pdf exec pdftocairo -pdf -f $page -l $page -x $xtl -y $ytl -W $w -H $h \ -paperw $w -paperh $h "$file" - | pdfcrop --margin $margin - fig$n.pdf Unlike Peter (and thanks to piet and haui), I can show an actual vector screenshot made by this script: The size of this shot is 2.7 kB. A bitmap of this size showing the same section is so terribly ugly that I've decided not to present it here.