ImageHash 4.3.1

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

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.

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!)

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