First we need to make sure we have the necessary packages installed.

apt-get install tesseract-ocr sane-utils

The tesseract-ocr package gives us a utility called tesseract which takes a TIFF file as input and will output the OCR’d .txt file of the tiff.

tesseract my.tif output

Now we need a command line method to grab the TIFF, sane-utils comes to the rescue. The command “scanimage” from sane will let us do that. It is a great little utility that I recommend reading up on to learn more about its features and options, as they may vary based on the type of scanner you have. My scanner has an Auto Document Feeder (ADF) so be aware that my instructions are specific to an ADF scanner.

This example is for scanning a letter sized piece of paper in batch mode saving output in the format of a TIFF

scanimage -y 279.4 -x 215.9 --batch --format=tiff --mode Lineart --resolution 300 --source ADF

This will output a new TIFF for each page that is scanned.

The below script combines several steps to output a single PDF document and .txt file for a scan job.

outname=$1
startdir=$(pwd)
tmpdir=scan-$RANDOM
 
cd /tmp
mkdir $tmpdir
cd $tmpdir
echo "################## Scanning ###################"
scanimage -y 279.4 -x 215.9 --batch --format=tiff --mode Lineart --resolution 300 --source ADF
echo "################### OCRing ####################"
i=1
for page in $(ls *.tif); do
        echo -n "Page: $i - "
        #run tesseract on each page and combine the outputs in a single file with a .txt extension.
        tesseract $page $page
        echo "---BEGIN PAGE: $i ---" >> $outname.txt
        cat $page.txt >> $outname.txt
        echo "---END PAGE: $1 ---" >> $outname.txt
        i=$(expr $i + 1)
done
mv $outname.txt $startdir
echo "############## Converting to PDF ##############"
#Use tiffcp to combine output tiffs to a single mult-page tiff
tiffcp -c lzw out*.tif output.tif 
#Convert the tiff to PDF
tiff2pdf output.tif > $startdir/$outname
cd ..
echo "################ Cleaning Up ################"
rm -rf $tmpdir
cd $startdir

I name the above script “scandoc” and it can be run by typing “scandoc myoutput.pdf” which will drop a pdf file (called myoutput.pdf) and a .txt (called myoutput.pdf.txt) file in the current directory with all the pages from the ADF. Very handy!

EDIT:
I’ve added Joe’s contributions in the comments to a gist at github.