OCR of a scanned PDF, using imagemagick, ghostscript, poppler and tesseract

pdfocr.sh

#!/bin/bash

# builds on: 
# https://ubuntuforums.org/showthread.php?t=1370827

# Prerequisites:
# On Mac OS:
# brew install --with-libtiff --with-ghostscript imagemagick
# brew install --all-languages tesseract
# brew install poppler

if [[ -z "$1" ]]
then
	echo "Usage: $0 files_to_OCR
	Examples:
	$0 file1 file2 ...
	"
	exit
fi

# choose language
read -p "Language of the PDFs [eng]: " LANG
LANG=${LANG:-eng}
if [ -z $(tesseract --list-langs | grep $LANG) ]; then
	echo "Language with code '$LANG' is not supported"
	exit 1
fi


TMPDIR=$(dirname $(mktemp -u))

for i in $*
do
	# check file
	if [[ ! -f "$i" ]]
	then
		echo "$i: file not found" >&2
		exit 1;
	fi
	
	# check if graphics file can be converted
	if ! identify "$i" 2>/dev/null 1>&2	
	then
		echo "$i: not convertable" >&2
		exit 1;
	fi

	echo "Processing '$i'" >&2
	
	#get number of pages and perform conversion one page at a time
	PAGES=$( pdfinfo $i | grep 'Pages' - | awk '{print $2}' )
	OUTPUT=$(echo $i| cut -d . -f1)-ocr.pdf
	PREFIX=$TMPDIR/ocrpdfpage
	for j in `seq 1 $PAGES`; do
		echo "Page $j"
		TMPFILE=$PREFIX$(printf %05d $j)
		# converting the graphic file
		echo ">>> converting to image"
		convert -monochrome -density 600 $i\[$(($j - 1 ))\] $TMPFILE.tiff
		# apply OCR on this file
		echo ">>> applying OCR"
		tesseract $TMPFILE.tiff -l $LANG $TMPFILE pdf >/dev/null 2>&1 # this suppresses stdout & stderr, comment out for debugging
		# remove temporary files
		rm $TMPFILE.tiff
	done
	# join temporary files
	echo "Joining pages...."
	gs -dBATCH -dNOPAUSE -q -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -sOutputFile=$OUTPUT $(ls $PREFIX*)
	# alternative : pdftk page_*.pdf cat output merged.pdf
	rm $PREFIX*
	echo "$i completed"
done

Unfortunately, on line 24 "tesseract --list-langs" sends its output to STDERR, breaking the pipe. Change this line to:

tesseract --list-langs 2>&1 >/dev/null | grep $LANG

Otherwise, this was a very useful script. Thanks!