Free image tools

Generate an image hash and perceptual fingerprint

Get two fingerprints for any picture: a SHA-256 checksum that identifies the exact file, and a perceptual hash that identifies what the picture looks like. Compare two images to measure how similar they are.

This tool needs JavaScript. It runs entirely in your browser; nothing is uploaded.

Private by design. Your files are processed by your own browser on your own device. They are not uploaded, stored or seen by Archive.London. Tool terms

How to use the Image Hash and Fingerprint tool

  1. Choose an image or drop it on the page.
  2. Copy the SHA-256 value to record the exact identity of the file.
  3. Note the perceptual hash (dHash and aHash) that describes the picture's appearance.
  4. Add a second image to compare the two and read the similarity result.

To generate an image hash, open the Image Hash and Fingerprint tool on Archive.London and drop in a file. It calculates the SHA-256 of the file for exact identity and a perceptual hash (dHash and aHash) of the picture for near-duplicate detection. Add a second image to see a similarity score. Everything is computed in your browser.

Two kinds of hash for two different questions

SHA-256 answers is this the identical file? It is a cryptographic hash of every byte, written as 64 hexadecimal characters. Change one bit, whether a pixel, a metadata field or a single character in a caption, and the result is completely different. Two files with the same SHA-256 can be treated as the same file.

Perceptual hashes answer is this the same picture? The tool reduces the image to a tiny greyscale grid and records a pattern of bits. An average hash (aHash) marks each cell as brighter or darker than the mean. A difference hash (dHash) marks whether each cell is brighter than its neighbour. Resizing, recompressing or converting the format changes few of those bits, so near-duplicates produce nearly identical hashes.

Use SHA-256 for integrity and identity. Use the perceptual hash to find the same image in different sizes and formats.

How similarity is measured, and where it fails

Two perceptual hashes are compared by counting the bits that differ, known as the Hamming distance. Few differing bits means the pictures look alike at a coarse level, and the comparison reports this as a similarity figure.

These simple hashes are sturdy against changes that leave the overall composition intact: scaling, JPEG recompression, mild brightness or colour shifts and format conversion. They are weak against changes that move the content around. A crop, a rotation, a mirror flip or a large added border can produce a very different hash from the same photograph.

False matches occur too. Images with little structure, such as a plain sky or a product on a white background, can score as similar when they are different frames. Treat a high score as a prompt to look at both pictures, not as a verdict.

Private by design: hashing happens on your device

Both hashes are calculated in your browser. The SHA-256 is computed over the file's bytes exactly as they are on disk, and the perceptual hash from the decoded picture. The image is not uploaded or stored and no account is needed, so the tool is suitable for unpublished or confidential material.

Because the SHA-256 covers the whole file, it will change if metadata is edited even though the picture looks the same. That is a feature when you need proof that a delivered file is byte-for-byte what was sent. If you want a fingerprint that survives a metadata edit, record the perceptual hash as well.

The usual browser limits apply: very large images depend on device memory, and files over about 100 megapixels may fail on a phone.

Checksums in digitisation and provenance logs

In a scanning project, a checksum is recorded for each master file when it is created. Recalculating it later and getting the same value shows that the file has not been corrupted or altered in storage or transfer, a routine known as fixity checking. Recorded with a date, the same value is a simple piece of provenance evidence that a particular file existed in a particular state.

Perceptual hashing addresses a different archive problem, duplicates. The Robert Harper Photograph Library inventory counts 95,487 unique photographs but 177,143 frames once duplicates and variants are included, as set out in the archive facts. Organisations with experience of this kind of work can read about the project on the digitisation partners page.

Frequently asked questions

What is the difference between SHA-256 and a perceptual hash?

SHA-256 is calculated from the file's bytes and changes completely if anything in the file changes, including metadata. It identifies an exact file. A perceptual hash is calculated from the picture's appearance and stays nearly the same when the image is resized, recompressed or converted. It identifies a picture across different copies. The two are complementary and this tool gives you both.

Can I use this to find out whether two photos are duplicates?

Yes. Load both images and the tool compares their perceptual hashes and reports how similar they are. Identical SHA-256 values mean the files are exact copies. A high perceptual similarity with different SHA-256 values usually means the same picture saved at another size, quality or format. Check visually when it matters, since plain or similar-looking images can score highly.

Will the hash change if I rename the file?

No. Neither hash uses the filename or the dates held by your operating system. The SHA-256 depends only on the contents of the file, and the perceptual hash only on the picture. Editing embedded metadata such as a caption or copyright field does change the SHA-256, because that data is stored inside the file, but it leaves the perceptual hash as it was.

Does a SHA-256 hash prove I own a photograph?

No. A hash shows that a specific file existed in a specific form, and a dated record of it can support a timeline. It says nothing about who created the picture or who holds the rights. Ownership is established by other evidence, such as original negatives or raw files, contracts and publication history. How much weight a hash record carries depends on the circumstances and the jurisdiction.

Why do a cropped photo and its original show as different?

Simple perceptual hashes divide the whole frame into a fixed grid, so cropping, rotating or flipping shifts the content into different cells and changes many bits. The method is designed to catch resized and recompressed copies, not rearranged ones. To compare a crop with its source, crop the original to the same area first and then hash both versions.

Licensing the archive for AI and data projects

95,487 photographs from one photographer and one rights holder. Read the dataset card, search the 503-folder inventory, discuss a sample set under NDA, or arrange a first look as scanning proceeds.

See AI and data licensingWrite to the archive