ocr.sh

#!/bin/sh
# This script turns ugly scanned images into plaintext
#
# @version 1.00 (ultra alpha)

# OCR program path; we assume tesseract, though.
OCR=tesseract
# Imagemagick
CONVERT=convert
# Do unpaper. This operation is costly, but can improve OCR.
UNPAPER=unpaper
UNPAPEROPTS=" --overwrite "
# Set this to 1 if you want more accurate OCR.
DOUNPAPER=0
# EXT, for easy cleanup
EXT=".OCR"

# SUPER dangerous, make $EXT very unique! Set to 1 to enable
# Edit: functionality removed.
DOCLEANUP=0

USAGE="Usage: ocr <outputfile> <filestobeconverted>"
ERR2="Unable to create output file."

# We need a minimum of two arguements
if [ $# -le 1 ]; then
   echo $USAGE
   exit 1
fi

OUTPUTFILE=$1

# Attempt to output some file. Run rm first to delete file if it exists,
# USERS BEWARE!
rm $OUTPUTFILE
touch $OUTPUTFILE

# We couldn't
if [ $? -ne 0 ]; then
   echo $ERR2
   exit 2
fi

for image in $*; do
   # Skip the first arguement....so ugly, clean when time permits.
   if [ $image == $OUTPUTFILE ]; then
      continue;
   fi
   
   # Do unpaper step
   if [ $DOUNPAPER == 1 ]; then
      $CONVERT $image $image$EXT.pnm
      $UNPAPER $UNPAPEROPTS $image$EXT.pnm $image$EXT.unpapered.pnm
      # tif required for tesseract
      $CONVERT $image$EXT.unpapered.pnm $image$EXT.tif
   # Convert to tif
   else 
      $CONVERT $image $image$EXT.tif   
   fi
   
   $OCR $image$EXT.tif $image$EXT
   # Make all new lines spaces. All lines end at the right of 
   # the page, so tesseract happily inserts a line break there
   # which we don't want. Now our doument fits on one line, 
   # awesome
   cat $image$EXT.txt | tr '\n' ' ' >> $OUTPUTFILE
   echo "---" >> $OUTPUTFILE
done;

echo "Finished ripping strings from images. Do rm *.$EXT.* to cleanup."