Extracts images from PDF while preserving PNG transparency

extract-pdf-images

#!/usr/bin/env bash

# Extracts images from a PDF into a directory, preserving transparency.
#
# Inspired by https://gist.github.com/bendavis78/ed22a974c2b4534305eabb2522956359
#
# Options:
#
# --help, -h       Display help
# --pdf, -p        The PDF to extract images from
# --directory, -d  The directory to output images to
#
# Examples:
#
# extract-pdf-images -p ./foo.pdf -d ./foo
#

# Exit on first error
set -e

# Debug
#set -x

# Checks that a file is valid
validfile() { [[ -r ${1} ]]; }

validdir() { [[ -w ${1} ]]; }

# Checks that a command is installed
installed () { command -v "${1}" >/dev/null 2>&1 || { >&2 echo "Cannot execute the ${1} command"; exit 1; } }

help() {
  scriptName=$(basename "$0")
  echo "${scriptName} [options]"
  echo
  echo Options:
  echo
  echo "--help, -h        Prints this help menu"
  echo "--pdf, -p         The PDF file to extract images from"
  echo "--directory, -d   The directory to extract images into"
  echo
  echo Example:
  echo
  echo "${scriptName}" --pdf "./foo.pdf" --directory "./foo"
  echo
}

installed pdfimages # Install poppler-utils to get pdfimages
installed convert

pdf=
outputDir=

# Exit if no parameters were provided
if [[ $# = 0 ]]; then
    help
    exit 1
fi

# Parse the parameters
while [[ $# != 0 ]]; do
    case "${1}" in
        -h|--help)
            help
            exit 0
            ;;
        -p|--pdf)
            pdf="$2"
            shift 2
            ;;
        -d|--directory)
            outputDir="${2}"
            shift 2
            ;;
        *)
            shift
            ;;
    esac
done

# Accept pdf name on stdin too
if [[ -z $pdf ]]; then
    pdf=$(cat)
fi

# Check the parameters
if ! validfile "${pdf}"; then
  >&2 echo "${pdf} cannot be read"
  exit 1
fi

if ! validdir "${outputDir}"; then
  >&2 echo "${outputDir} cannot be written to"
  exit 1
fi

tmpdir=$(mktemp -d /tmp/extract-pdf-images-XXXXXX)
mkdir "${tmpdir}/extracted"

# Remove the tmpdir when we're done
trap 'rm -rf ${tmpdir}' EXIT INT

pdfimages -all "${pdf}" "${tmpdir}/extracted/image" || exit 1

# Rename images based on object id and whether or not they are a mask
pdfimages -list "${pdf}" | tail -n +3 | while read -r row; do
    num=$(echo "$row" | awk '{print $2}');
    imgtype=$(echo "$row" | awk '{print $3}');
    imgenc=$(echo "$row" | awk '{print $9}');
    objectid=$(echo "$row" | awk '{print $11}');

    if [[ "$imgenc" == "jpeg" ]]; then
        ext="jpg";
    else
        ext="png";
    fi
    src=$(printf "${tmpdir}/extracted/image-%03d.$ext" "$num");

    if [[ "$imgtype" == "smask" ]]; then
        dest=$(printf "${tmpdir}/image-%03d-mask.$ext" "$objectid");
    else
        dest=$(printf "${tmpdir}/image-%03d.$ext" "$objectid");
    fi
    #echo "$src -> $dest";
    mv "$src" "$dest" || exit 1;
done

rmdir "${tmpdir}/extracted"

## Merge the images that have a mask
pdfimages -list "${pdf}" | tail -n +3 | while read -r row; do
    imgtype=$(echo "$row" | awk '{print $3}');
    imgenc=$(echo "$row" | awk '{print $9}');
    objectid=$(echo "$row" | awk '{print $11}');

    if [[ "$imgenc" == "jpeg" ]]; then
        ext="jpg";
    else
        ext="png";
    fi

    if [[ "$imgtype" == "smask" ]]; then
        mask=$(printf "${tmpdir}/image-%03d-mask.${ext}" "$objectid");

        # Use the image that corresponds to this mask. Sometimes the image and its mask are different types.
        pngimg=$(printf "${tmpdir}/image-%03d.png" "$objectid");
        jpgimg=$(printf "${tmpdir}/image-%03d.jpg" "$objectid");
        if validfile "${pngimg}"; then
          img=$pngimg
        else
          img=$jpgimg
        fi

        #echo "convert $img $mask";
        convert "$img" "$mask" -alpha off -compose copy-opacity -composite "$img" || exit 1;
    fi
done

rm -f "${tmpdir}"/image-*-mask.*;
mv "${tmpdir}"/* "${outputDir}/";