Go implementation of multiple perceptual hash algorithms for images.

Documentation

Detailed documentation has moved to the Wiki:

• Wiki Home
• Algorithms
• Similarity Metrics
• Convenience Functions
• Interpolation Methods
• Migration Guide

Installing

go get -u github.com/ajdnik/imghash/v2

import "github.com/ajdnik/imghash/v2"

Most consumers only need the top-level imghash package. Core types (Hash, Binary, UInt8, Float64, Distance) are re-exported there.

Quick Start

If you're unsure which hash to pick, start with PDQ.

package main

import (
  "fmt"

  "github.com/ajdnik/imghash/v2"
)

func main() {
  pdq, err := imghash.NewPDQ()
  if err != nil {
    panic(err)
  }

  h1, err := imghash.HashFile(pdq, "image1.png")
  if err != nil {
    panic(err)
  }

  h2, err := imghash.HashFile(pdq, "image2.png")
  if err != nil {
    panic(err)
  }

  dist, err := pdq.Compare(h1, h2)
  if err != nil {
    panic(err)
  }

  fmt.Printf("Distance: %v\n", dist)
}

Algorithms at a Glance

Algorithm	Hash type	Default metric
Average	Binary	Hamming
Difference	Binary	Hamming
Median	Binary	Hamming
PHash	Binary	Weighted Hamming
WHash	Binary	Hamming
MarrHildreth	Binary	Hamming
BlockMean	Binary	Hamming
PDQ	Binary	Hamming
RASH	Binary	Hamming
ColorMoment	Float64	L2 (Euclidean)
Zernike	Float64	L2 (Euclidean)
GIST	Float64	Cosine
BoVW (Histogram)	Float64	Cosine
BoVW (MinHash)	Float64	Jaccard
BoVW (SimHash)	Binary	Jaccard
CLD	UInt8	L2 (Euclidean)
EHD	UInt8	L1 (Manhattan)
LBP	UInt8	Chi-Square
HOGHash	UInt8	Cosine
RadialVariance	UInt8	L1 (Manhattan)

See Algorithms for options, defaults, references, and examples.

Community

• Code of Conduct
• Contributing Guide
• Security Policy
• Migration Guide

License

Imghash is released under the MIT license. See LICENSE.