aHash, dHash, pHash
Three complementary perceptual hashes: average, gradient and cosine transform (DCT).
Analysis
Image compare
The Image compare module uses perceptual hashes to tell whether two images are the same scene (perhaps recompressed) or a manipulation, documenting everything in a side-by-side PDF report.
What it does
What the Image compare module does
Distance and similarity
Hamming distance between the hashes and a similarity percentage for each algorithm.
Threshold verdict
Identical / similar / different based on adjustable thresholds (defaults 90% and 70%).
Side-by-side report
Landscape PDF with the two images side by side, the distances and the metadata/hash differences.
Step by step
How it works
- Load the two images and choose the hash depth (64/144/256 bits).
- Probatio computes aHash, dHash and pHash on downscaled grayscale versions.
- It compares with the Hamming distance, issues the verdict and generates the PDF report.
FAQ
Frequently asked questions
What is a perceptual hash?
It is a fingerprint that represents an image's appearance: two similar images have close perceptual hashes, even if the bytes differ.
How is similarity measured?
With the Hamming distance between the hashes, converted to a percentage; the verdict (identical/similar/different) uses adjustable thresholds.
Does it help detect deepfakes?
It helps compare versions of an image and document differences; for capture authenticity use the C2PA analysis in the Image analysis module.
Related modules