Remove redaction blobs from PDF before OCR

oneliner.sh

set -x
set -e

mkdir -p output

cd original
for f in *.jpg; do
  # first image in the stack is the base image
  convert $f \
    # second image is the same thing, but with dilate/erode applied to remove thin lines (i.e., text)
    # all that remains are blobs and larger shapes
    \( $f -morphology Dilate Octagon:2 -morphology Erode Octagon:3 \) \
    # compose the two together using subtraction
    # the subtracted image is white-on-black, so negate to reverse it back
    -compose Minus -composite -negate ../output/$f
done
cd ..

stages.sh

# this version has step-by-step output for debugging
set -e
set -x

# generate page images
mkdir -p original
pdftocairo -jpeg original.pdf original/page

# dilate/erode to just get blob images
mkdir -p 2_eroded
cd original
for f in *.jpg; do
  convert $f -morphology Dilate Octagon:2 -morphology Erode Octagon:3 ../2_eroded/$f
done
cd ..

# subtract the blobs from the original
mkdir -p 3_subtracted
cd original
for f in *.jpg; do
  composite -compose Minus ../2_eroded/$f $f ../3_subtracted/$f
done
cd ..

# convert back from negative form
mkdir -p 4_reverted
cd 3_subtracted
for f in *.jpg; do
  convert $f -negate ../4_reverted/$f
done