Recently I wanted to print a few scanned pages. However, due to the low quality of the source material (see below) reducing the pages to black&white didn't exactly improve them.

Enter unpaper. While unpaper is a rather unknown tool it's also extremely useful because it allows diverse image modifications to improve scanned pages, e.g.:

• black-/gray-/noise-/blurfiltering
• deskewing
• border-aligning
• mask-centering
• etc.

If you aren't sold yet the documentation offers other great examples of applied image processing techniques.

Below I've included the minimal¹⁾ script I used to process the scanned pages. While the settings produced sufficient results for my needs they are probably still far from perfect – so don't forget to toy with your settings to get optimal results ,)

#!/bin/bash
# Usage: autounpaper *.jpg
 
inputfiles=$*
tmpin=`mktemp -t autounpaper-in-XXXXXXXXXX`.pgm
tmpout=`mktemp -t autounpaper-out-XXXXXXXXXX`.pgm
 
opts="-q --overwrite"
opts+=" --no-deskew --no-mask-scan"
opts+=" --grayfilter-size 1,1 --grayfilter-step 1,1 --grayfilter-threshold 0.4"
 
echo -n "Processing:"
 
for input in $inputfiles; do
	echo -n " $input"
	ending=${input#*.}
	output=${input%.*}-unpaper.$ending
 
	convert $input $tmpin && echo -n . && \
	unpaper $opts $tmpin $tmpout && echo -n . && \
	convert $tmpout $output
done
 
rm $tmpin $tmpout
 
echo