This is an implementation of a perceptual image hash,
Block Mean Value Based Image Perceptual Hashing and discussed in this StackOverflow question.
I found the phash package to be a little error prone with respect to file I/O. This package, while the API is very similar, is different in some key ways.
phashbinds directly to the pHash library.
canvas-phashis a direct implementation, written in coffeescript.
phashis callback-based while
canvas-phashis promise-based (specifically, it uses bluebird for promise management).
phashgenerally takes longer to compute the hash of an image but is faster at finding the hamming distance between two hashes.
- The hash output by
phashis an integer, expressed as a string. The hash output by
canvas-phashis a 128-byte
- Comparing the two libraries on the basis of the correlation between hamming distance and "perceived difference" had mixed results.
phashwas better at some things,
canvas-phashwas better at others.
I ran some preliminary tests to check the performance against
phash and found it's fairly comparable.
Computing A Hash
The time taken ranged from just under 75ms to 150ms. For my tests, it generally took
phash about 1-2 times longer to compute a hash as it took
Finding the Hamming Distance
Typical time taken ranged from 0.2ms to 0.3ms. For my tests, it generally took
canvas-phash about 2-3 times longer to find the hamming distance of two hashes. When comparing against a large collection of images, this is potentially significant. That being said, this library has not been optimized. Also, the actual hash created is 128 bytes long and takes up about 2-3 times more space.
getImageHash- Accepts 1 parameter, the path of the image. Returns a promise with eventual value equal to the "Block Mean Value Based" pHash.
getHammingDistance- Accepts 2 parameters, two instances of
Bufferof length 128 (this is what is returned from
getSHA256- This computes the SHA256 hash of the pixel data. The only parameter is setup like that of
getImageHash. This is useful for fast checks of exact matches. Ignores metadata.
readImage- Reads an image at the specified path and returns an object with properties:
data, the byte array,
width, the width of the image, and
height, the height of the image.
(Another example exists in the repo)
phash = require 'canvas-phash'Promise = require 'bluebird'PromiseallphashgetImageHash 'image.jpg'phashgetImageHash 'otherImage.jpg'spreaddist = phashgetHammingDistance hash1hash2
In the previous example, Promise.all is used to make the code readable.
bluebird is not necessary to use this package. The typical use-case would be to compute the hash of a single image via
phash.getImageHash('image.jpg').then (hash)-> and compare that against a list of pre-existing hashes for close matches.