ImageHash 4.3.1

Image Hashing library

An image hashing library written in Python. ImageHash supports:

• Average hashing

• Perceptual hashing

• Difference hashing

• Wavelet hashing

• HSV color hashing (colorhash)

• Crop-resistant hashing

Rationale

Image hashes tell whether two images look nearly identical. This is different from cryptographic hashing algorithms (like MD5, SHA-1) where tiny changes in the image give completely different hashes. In image fingerprinting, we actually want our similar inputs to have similar output hashes as well.

The image hash algorithms (average, perceptual, difference, wavelet) analyse the image structure on luminance (without color information). The color hash algorithm analyses the color distribution and black & gray fractions (without position information).

Installation

Based on PIL/Pillow Image, numpy and scipy.fftpack (for pHash) Easy installation through pypi:

pip install imagehash

Basic usage

>>> from PIL import Image
>>> import imagehash
>>> hash = imagehash.average_hash(Image.open('tests/data/imagehash.png'))
>>> print(hash)
ffd7918181c9ffff
>>> otherhash = imagehash.average_hash(Image.open('tests/data/peppers.png'))
>>> print(otherhash)
9f172786e71f1e00
>>> print(hash == otherhash)
False
>>> print(hash - otherhash)  # hamming distance
33

Each algorithm can also have its hash size adjusted (or in the case of colorhash, its binbits). Increasing the hash size allows an algorithm to store more detail in its hash, increasing its sensitivity to changes in detail.

The demo script find_similar_images illustrates how to find similar images in a directory.

Source hosted at GitHub: https://github.com/JohannesBuchner/imagehash

References

• Average hashing (aHashref)

• Perceptual hashing (pHashref)

• Difference hashing (dHashref)

• Wavelet hashing (wHashref)

• Crop-resistant hashing (crop_resistant_hashref)

Examples

To help evaluate how different hashing algorithms behave, below are a few hashes applied to two datasets. This will let you know what images an algorithm thinks are basically identical.

Example 1: Icon dataset

Source: 7441 free icons on GitHub (see examples/github-urls.txt).

The following pages show groups of images with the same hash (the hashing method sees them as the same).

• phash (or with z-transform)

• dhash (or with z-transform)

• colorhash

• average_hash (with z-transform)

The hashes use hashsize=8; colorhash uses binbits=3. You may want to adjust the hashsize or require some manhattan distance (hash1 - hash2 < threshold).

Example 2: Art dataset

Source: 109259 art pieces from http://parismuseescollections.paris.fr/en/recherche/image-libre/.

The following pages show groups of images with the same hash (the hashing method sees them as the same).

• phash (or with z-transform)

• dhash (or with z-transform)

• colorhash

• average_hash (with z-transform)

For understanding hash distances, check out these excellent blog posts: * https://tech.okcupid.com/evaluating-perceptual-image-hashes-at-okcupid-e98a3e74aa3a * https://content-blockchain.org/research/testing-different-image-hash-functions/

Storing hashes

As illustrated above, hashes can be turned into strings. The strings can be turned back into a ImageHash object as follows.

For single perceptual hashes:

>>> original_hash = imagehash.phash(Image.open('tests/data/imagehash.png'))
>>> hash_as_str = str(original_hash)
>>> print(hash_as_str)
ffd7918181c9ffff
>>> restored_hash = imagehash.hex_to_hash(hash_as_str)
>>> print(restored_hash)
ffd7918181c9ffff
>>> assert restored_hash == original_hash
>>> assert str(restored_hash) == hash_as_str

For crop_resistant_hash:

>>> original_hash = imagehash.crop_resistant_hash(Image.open('tests/data/imagehash.png'), min_segment_size=500, segmentation_image_size=1000)
>>> hash_as_str = str(original_hash)
>>> restored_hash = imagehash.hex_to_multihash(hash_as_str)
>>> assert restored_hash == original_hash
>>> assert str(restored_hash) == hash_as_str

For colorhash:

>>> original_hash = imagehash.colorhash(Image.open('tests/data/imagehash.png'), binbits=3)
>>> hash_as_str = str(original_hash)
>>> restored_hash = imagehash.hex_to_flathash(hash_as_str, hashsize=3)

For storing the hashes in a database and using fast hamming distance searches, see pointers at https://github.com/JohannesBuchner/imagehash/issues/127 (a blog post on how to do this would be a great contribution!)

Changelog

• 4.3: typing annotations by @Avasam @SpangleLabs and @nh2

• 4.2: Cropping-Resistant image hashing added by @joshcoales

• 4.1: Add examples and colorhash

• 4.0: Changed binary to hex implementation, because the previous one was broken for various hash sizes. This change breaks compatibility to previously stored hashes; to convert them from the old encoding, use the “old_hex_to_hash” function.

• 3.5: Image data handling speed-up

• 3.2: whash now also handles smaller-than-hash images

• 3.0: dhash had a bug: It computed pixel differences vertically, not horizontally.

  I modified it to follow dHashref. The old function is available as dhash_vertical.

• 2.0: Added whash

• 1.0: Initial ahash, dhash, phash implementations.

Contributing

Pull requests and new features are warmly welcome.

If you encounter a bug or have a question, please open a GitHub issue. You can also try Stack Overflow.

Other projects

• http://blockhash.io/

• https://github.com/acoomans/instagram-filters

• https://pippy360.github.io/transformationInvariantImageSearch/

• https://www.phash.org/

• https://pypi.org/project/dhash/

• https://github.com/thorn-oss/perception (based on imagehash code, depends on opencv)

• https://docs.opencv.org/3.4/d4/d93/group__img__hash.html