Reference Glossary
File hash / fingerprint
A digest that identifies a file's content — either cryptographically exact (MD5/SHA) or perceptually similar (pHash) — used for exact- or near-duplicate detection in DAM libraries.
Why it matters in a DAM
A cryptographic hash only matches identical bytes, so it can't tell a DAM that two exports of the same photo — one resized, one re-saved as a different format, one with a watermark added — are related; a perceptual fingerprint compares the visual content itself and stays similar across those changes, which is what powers a "find similar or duplicate" feature in a DAM search. That distinction matters because photographers and designers routinely upload the same shoot exported at multiple resolutions and formats, and without perceptual fingerprinting the DAM sees them as entirely unrelated files.
A worked example
Common mistake
Teams rely solely on cryptographic hashing for duplicate detection and miss near-duplicates like recompressed, resized, or lightly edited copies of the same asset, because catching those requires perceptual hashing layered on top of cryptographic hashing — and many DAM platforms implement only the exact-match kind.
“File hash” and “fingerprint” are sometimes used interchangeably, but they describe two different techniques with different guarantees. A cryptographic hash — MD5, SHA-1, SHA-256 — is an exact-match technique: it produces a completely different value for any change to the underlying bytes, which makes it strict but blind to visual similarity. A perceptual fingerprint (commonly called pHash) instead derives a signature from the image’s visual features, so two images that look nearly identical produce fingerprints that are numerically close, measured by something like Hamming distance, even if their underlying bytes differ entirely.
In a DAM, both are useful for different jobs. Cryptographic hashing catches the case of the exact same file uploaded twice under different names — a genuine storage-saving deduplication. Perceptual fingerprinting catches the more common real-world case: the same photo exported at a different resolution, re-saved through a different tool, or lightly cropped, which cryptographic hashing sees as completely unrelated but a human reviewer would immediately recognize as the same shot.
Perceptual fingerprinting is not free of trade-offs — it is well suited to catching exact and mildly modified duplicates but degrades on heavily transformed or stylistically altered near-duplicates, so it complements rather than replaces manual review or cryptographic hashing, and a DAM that only implements one of the two will systematically miss one category of duplicate.
Frequently asked
What's the difference between a cryptographic hash and a perceptual fingerprint?
A cryptographic hash like MD5 or SHA-256 is an exact-match technique — any byte change produces a totally different value. A perceptual fingerprint (pHash) derives a signature from an image's visual features, so visually similar images produce numerically close fingerprints even when the underlying bytes differ entirely.
Why do DAM libraries need both hash types?
Cryptographic hashing catches exact-duplicate uploads of the same file under a different name. Perceptual fingerprinting catches the more common case of the same photo re-exported at a different resolution, re-saved through a different tool, or lightly cropped.
How is fingerprint similarity typically measured?
By Hamming distance between the perceptual hash bits — the smaller the distance, the more visually similar the two images are considered.
What duplicates does relying only on cryptographic hashing miss?
Near-duplicates like recompressed, resized, or lightly edited copies of the same asset. Catching those requires perceptual hashing layered on top, and many DAM platforms only implement the exact-match kind.
Is perceptual fingerprinting reliable for every kind of edit?
No. It's well suited to catching exact and mildly modified duplicates but degrades on heavily transformed or stylistically altered near-duplicates, so it complements rather than replaces manual review or cryptographic hashing.
Give a real-world DAM scenario where perceptual fingerprinting matters.
Photographers and designers routinely upload the same shoot exported at multiple resolutions and formats. Without perceptual fingerprinting, the DAM sees these as entirely unrelated files.
Sources
- A perceptual hash, unlike cryptographic hashes such as MD5 or SHA1, is designed so that hashes of visually similar images are numerically close, typically measured with Hamming distance between the hash bits. checked 2026-08-07 — Ben Hoyt — Duplicate image detection with perceptual hashing in Python