Skip to content

Perceptual Image Hashes

Perceptual Image Hashing refers to the use of a fingerprint algorithm to generate a hash that correlates with the appearance of an image without depending on other parameters such as file size, type, name, resolution, or other metadata.

Sorting by Visual Similarity

PhotoPrism currently generates a perceptual hash while indexing, which can be used to sort search results by visual similarity, e.g.:

Note, though, that it is not yet possible to automatically stack files based on this and that the hash is not 100% precise, so visually different images may have the same hash. This is because it was developed for sorting and not for stacking or finding duplicates.

Stacking of Similar Files

Developing a user-friendly web interface for semi-automatic stacking of a large number of images based on their appearance is expected to require a significant amount of work.

A simpler solution could be to add one or more additional stacking options on the library settings page and allow them to be applied to existing pictures by pressing a button or running a terminal command, but without being able to see the result in advance or perform manual changes.

See the following issues for related feature requests:

Software Libraries & Examples

References, Algorithms & Tutorials