Extracts images from PDF while preserving PNG transparency

extract_pdf_images.sh

#!/bin/bash
usage() {
    echo "Usage: $(basename $0) in.pdf dest";
}

[[ -z "$1" ]] && usage && exit 1;
[[ -z "$2" ]] && usage && exit 1;

TMPDIR="$(mktemp -d)";
DIR=$2;

mkdir "$TMPDIR/extracted";
# Extract the images into tmpdir
pdfimages -all $1 "$TMPDIR/extracted/image" || exit 1;

# Rename images based on object id and whether or not they are a mask
pdfimages -list $1 | tail -n +3 | while read row; do
    num=$(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

# Merge the images that have a mask
pdfimages -list $1 | tail -n +3 | while read row; do
    imgtype=$(echo "$row" | awk '{print $3}');
    objectid=$(echo "$row" | awk '{print $11}');
    if [[ "$imgtype" == "smask" ]]; then
        img=$(printf "$TMPDIR/image-%03d.png" $objectid);
        mask=$(printf "$TMPDIR/image-%03d-mask.png" $objectid);
        echo "convert $img $mask";
        convert "$img" "$mask" -alpha off -compose copy-opacity -composite "$img" || exit 1;
    fi
done

rm "$TMPDIR"/image-*-mask.png*;

mv $TMPDIR/* "$DIR/";

Usefull, worked fine for me for a PDF with all images having transparent background, thanks for having shared this script

Thanks!

I found a PDF where the image and its mask are different file types. This script assumes they are the same.

Thanks for this script. It made a great starting point for a larger project I am tackling. I took your gist and rewrote it in Python to better handle image types and different composition modes.

https://gist.github.com/XBigTK13X/4796a0ca7f16e83438914384a57dc46b