Compares all images in a directory, doesn't check the same pair twice, duplicates are opened in Preview

image_compare.sh

#!/usr/bin/env zsh

# Threshold for similarity
THRESHOLD=0.05

# Get a sorted list of PNG files, using null-terminated strings to handle special characters
files=()
while IFS= read -r -d '' file; do
    files+=("$file")
done < <(find . -maxdepth 1 -type f -name "*.png" -print0 | sort -z)

# Find similar images in a directory
for ((i=1; i<=${#files}; i++)); do
    echo -n "."
    for ((j=i+1; j<=${#files}; j++)); do
        file1="${files[i]}"
        file2="${files[j]}"
        # Run comparison in background
        (
            # Use printf to safely escape filenames for the compare command
            similarity=$(compare -metric AE "${file1}" "${file2}" null: 2>&1)
            
            # Extract just the numeric value, handling both possible output formats
            similarity_float=$(echo "$similarity" | sed 's/[()]//g' | awk '{print $1}')
            
            # Check if similarity_float is actually a number and compare
            if [[ "$similarity_float" =~ ^[0-9]*\.?[0-9]+$ ]]; then
                result=$(echo "$similarity_float < $THRESHOLD" | bc -l 2>/dev/null)
                if [ "$result" -eq 1 ]; then
                    echo -n "#"
                    # Safely open files
                    open "${file1}" "${file2}"
                fi
            fi
        ) &
    done
done

wait
echo "\nComparison complete"

by Matt Sephton, March 2025, MIT