Extract slides from a presentation video by comparing the PSNR of consecutive difference of down-sampled frames against a threshold

extract-slides.sh

#!/bin/bash
set -e

# Usage: ./extract-slides.sh mylecture.mp4

# Config variables
SIMILARITY_THRESHOLD=25 # A slide change is assumed iff frames PSNR similarity falls below this value
FRAME_PERIOD=30 # [Seconds] Duration between sampling times
CONVERT_TO_PDF=0 # Set to 1 to make pdf. Might cause errors if imagemagick is not configured properly. See https://askubuntu.com/questions/1081895/trouble-with-batch-conversion-of-png-to-pdf-using-convert
PDF_FILE="${1%.*}-slides.pdf" # Only used if CONVERT_TO_PDF is 1
OUTPUT_PATH="${1%.*}-slides" # Only used if CONVERT_TO_PDF is 0
FRAMES_TEMP_DIRECTORY="/tmp/extract-slides/frames/${1%.*}"
SLIDES_TEMP_DIRECTORY="/tmp/extract-slides/slides/${1%.*}"

# Get directories ready
mkdir -p "$FRAMES_TEMP_DIRECTORY"
mkdir -p "$SLIDES_TEMP_DIRECTORY"
rm -f "$FRAMES_TEMP_DIRECTORY"/*.jpg
rm -f "$SLIDES_TEMP_DIRECTORY"/*.jpg

# Slice the video into frames
ffmpeg -i "$1" -vf fps=1/$FRAME_PERIOD "$FRAMES_TEMP_DIRECTORY/%04d.jpg"

# Detect slides
last_frame=NONE
current_slide_number=1
IFS=$'\n'
for frame in $(ls "$FRAMES_TEMP_DIRECTORY"/*)
do
    current_slide_name=$(printf "%04d.jpg" $current_slide_number)
    if [ "$last_frame" == "NONE" ]
    then
        cp "$frame" "$SLIDES_TEMP_DIRECTORY"/$current_slide_name
    else
        similarity=$(ffmpeg -i "$frame" -i "$last_frame" -filter_complex "psnr" -f null /dev/null 2>&1 | grep PSNR | grep -Po "average:\K(\d+|inf)")
        if [ "$similarity" != "inf" ] && [ "$similarity" -lt "$SIMILARITY_THRESHOLD" ]
        then
            # New slide
            current_slide_number=$(($current_slide_number+1))
            current_slide_name=$(printf "%04d.jpg" $current_slide_number)
            cp "$frame" "$SLIDES_TEMP_DIRECTORY/$current_slide_name"
        else
            # Old slide, unchanged or with small updates
            cp "$frame" "$SLIDES_TEMP_DIRECTORY/$current_slide_name"
        fi
    fi
    last_frame="$frame"
done

# Convert to pdf
if [ "$CONVERT_TO_PDF" == "1" ]
then
    convert $(ls "$SLIDES_TEMP_DIRECTORY"/*) "$PDF_FILE"
else
    mkdir -p "$OUTPUT_PATH"
    rm -f "$OUTPUT_PATH"/*.jpg
    cp "$SLIDES_TEMP_DIRECTORY"/* "$OUTPUT_PATH/"
fi

# Clear temp files
rm -rf "$FRAMES_TEMP_DIRECTORY"
rm -rf "$SLIDES_TEMP_DIRECTORY"