Bash script to convert image PDFs to text using OCR. The script uses pdftk, imagemagick, ghostscript and tesseract-ocr.

ocr.sh

#!/bin/sh

#needs: pdftk, imagemagick, ghostscript and tesseract-ocr
#pdf should be 300dpi or higher resolution

#set -o errexit

function success () {
  echo "\033[33;32m ✓ \033[33;00m"
}

function failure () {
  echo "\033[33;31m ✗"
  echo "\033[33;31mError: "$1"" 1>&2
  exit 1
}

LANG="eng"

if [ -z "$1" ]; then
  echo "Opening file ...\c"
  failure "No input file specified."
fi

if [ -z "$2" ]; then
  echo "No language specified, using English as default"
else
  LANG=$2
fi

echo "Splitting PDF file ...\c"
pdftk "$1" burst
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

echo "Converting PDF files to images ...\c"
trap "exit" INT
for i in pg*.pdf
  do convert -units pixelsperinch -density 300x300 -colorspace Gray -depth 8 $i "`basename $i .foooo`.png"
done;
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

echo "Converting images to monochrome ...\c"
trap "exit" INT
for i in pg*.png
  do convert $i +dither -monochrome -normalize "`basename $i .png`-m.png"
done;
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

echo "Character recognition ...\c"
trap "exit" INT
for i in *-m.png
  #change language here, ex. -l eng
  do tesseract $i "`basename $i .png`" -l $LANG quiet
done;
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

echo "Preparing text file ...\c"
DUMP=`basename "$1" .pdf`.txt;
touch $DUMP;
cat pg*.txt >> $DUMP;
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

echo "Cleaning up ...\c"
mkdir .trash;
mv pg* .trash;
rm -rf .trash
if [ $? -eq 0 ]; then
    success
else
    failure "$?"
fi

exit 0