OCR of scans with splitting double-page

OCR_PDF.sh

#!/bin/bash
for pdf in *.pdf;
do
	# Display the PDF file to process
	echo "Processing file \"$pdf\"…"

	# Temporary folder
	TmpRep="/tmp/conversion$(date +%Y%m%d%H%M%S)"
	mkdir --parents "$TmpRep"

	# Extract the PDF in the temporary folder
	pdfimages "$pdf" "$TmpRep/${pdf%.pdf}"

	for img in "$TmpRep"/*.ppm;
	do
		# Get image size
		Larg=$(identify -format "%w" "$img")
		Haut=$(identify -format "%h" "$img")

		# Check if the image is in landscape
		if [ $Larg -gt $Haut ]
		then
		  # Cut it in two portraits
		  echo "The image \"$img\" is in landscape. It will be cut in two portraits."
		  convert "$img" -crop 2x1@ "${img%.*}_%d.ppm"

		  # Delete the original
		  rm "$img"
		fi
	done

	# Create the PDF file with OCR from the images
	img2pdf "$TmpRep"/*.ppm | ocrmypdf --language $1 --deskew --remove-background --clean-final --tesseract-timeout 240 --optimize 3 - "${pdf%.pdf} - OCR.pdf"

	# Delete the temporary folder
	rm -rf "$TmpRep"/
	echo "File processing \"$pdf\" done."
done