Exact hash or visual fingerprint? Use the one that matches the job
Two images can look the same while having different file bytes. MAXScanner therefore keeps cryptographic SHA-256 separate from perceptual aHash, dHash and pHash instead of presenting one score as proof of duplication.
SHA-256 answers the exact-file question
SHA-256 is calculated from the complete file bytes. If two SHA-256 values match, the compared files are byte-for-byte identical for this workflow. A metadata edit, re-export, resize or recompression changes the bytes and normally changes SHA-256 even when the picture still looks the same.
Use the exact digest for integrity checks, transfer verification and exact duplicate detection. Keep the digest with the specific file version that was checked.
Perceptual hashes answer a different question
aHash summarizes average lightness, dHash summarizes neighboring pixel changes, and pHash uses low-frequency image structure. Each produces a compact 64-bit visual fingerprint after the image is decoded and normalized.
When two images have different SHA-256 values but small perceptual Hamming distances, they can be useful near-duplicate candidates to inspect. A resized, re-encoded or lightly adjusted image may remain visually close even though its exact file hash changes.
Read Hamming distance as a screening signal
Hamming distance counts how many fingerprint bits differ. A distance of 0 means the selected perceptual hash is identical; larger distances indicate more fingerprint disagreement. MAXScanner also shows a simple percentage derived from that bit distance for readability.
There is no universal threshold that proves two photographs are duplicates. Crops, borders, rotation, heavy color edits, text overlays and different image classes can change each algorithm differently, so the tool labels close matches conservatively and asks you to confirm them visually.
Everything stays in the browser
The image is decoded, sampled and hashed in the browser. This makes the tool useful for private QA, asset cleanup and pre-upload checks without sending the image bytes to a MAXScanner processing server.
For a large library, use this page to understand and verify the fingerprint logic before building any automated deletion workflow. Never delete originals solely from a perceptual score.
Do not use a perceptual score as a delete button
Treat close fingerprints as candidates for review. Keep originals until you have visually confirmed the relationship and checked whether metadata, crop, resolution or editing differences are important to your workflow.