Overview
Average Hash (also known as aHash) is a simple perceptual hashing algorithm described by Dr. Neal Krawetz in “Looks Like It”. It compares each pixel against the mean brightness to produce a binary hash.
This is one of the fastest and simplest perceptual hashing algorithms, making it ideal for quick similarity checks.
When to Use
Use Average Hash when you need:
- Fast computation with minimal overhead
- Simple implementation that’s easy to understand
- Basic similarity detection for images
- Low memory footprint (64-bit default hash)
- Quick duplicate detection in small to medium datasets
Average Hash is less robust to transformations than DCT-based methods like PDQ or PHash. Use it for basic similarity detection rather than precise matching.
Constructor
func NewAverage(opts ...AverageOption) (Average, error)
Available Options
WithSize(width, height uint) - Sets the resize dimensionsWithInterpolation(interp Interpolation) - Sets the resize interpolation methodWithDistance(fn DistanceFunc) - Overrides the default Hamming distance function
Supported Interpolation Methods
NearestNeighborBilinear (default)BicubicMitchellNetravaliLanczos2Lanczos3BilinearExact
Usage Example
package main
import (
"fmt"
"github.com/ajdnik/imghash/v2"
)
func main() {
// Create Average hasher with default settings
avg, err := imghash.NewAverage()
if err != nil {
panic(err)
}
// Hash an image file
hash, err := imghash.HashFile(avg, "image.jpg")
if err != nil {
panic(err)
}
fmt.Printf("Average hash: %v\n", hash)
}
With Custom Options
// Create Average hasher with custom size
avg, err := imghash.NewAverage(
imghash.WithSize(16, 16), // 256-bit hash
imghash.WithInterpolation(imghash.Lanczos3),
)
if err != nil {
panic(err)
}
hash, err := imghash.HashFile(avg, "image.jpg")
Default Settings
- Hash size: 64 bits (8 bytes)
- Resize dimensions: 8×8 pixels
- Interpolation: Bilinear
- Distance metric: Hamming distance
How It Works
The Average Hash algorithm:
- Resizes the image to the specified dimensions (default 8×8)
- Converts to grayscale
- Computes the mean pixel value
- For each pixel:
- Sets bit to 1 if pixel value > mean
- Sets bit to 0 if pixel value ≤ mean
- Produces a binary hash (width × height bits)
Comparison
Average hashes are compared using Hamming distance. Lower distances indicate more similar images.
avg, _ := imghash.NewAverage()
h1, _ := imghash.HashFile(avg, "image1.jpg")
h2, _ := imghash.HashFile(avg, "image2.jpg")
dist, err := avg.Compare(h1, h2)
if err != nil {
panic(err)
}
if dist < 10 {
fmt.Println("Images are similar")
}
For 8×8 hashes (64 bits), a Hamming distance of 0-10 typically indicates very similar images, 10-20 indicates some similarity, and >20 indicates different images.
- Simple resize operation
- Single-pass mean calculation
- Direct pixel comparison
- No complex transforms (DCT, DWT, etc.)
References