Perceptual Hash

Generate a perceptual hash for an image, which can be used for similarity comparison, duplicate detection, or content-based image retrieval.

Request Parameters

Response

Returns a JSON object containing the perceptual hash of the image in various formats.

Example Request

    // example.sh
curl -X POST "https://oyyi.xyz/api/image/p-hash" \
  -H "accept: application/json" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@image.jpg"
  

Example with Python

    // get_phash.py
import requests
import json

url = "https://oyyi.xyz/api/image/p-hash"

files = {
    'file': open('image.jpg', 'rb')
}

response = requests.post(url, files=files)

# Parse the JSON response
hash_data = response.json()
print(json.dumps(hash_data, indent=2))
  

Example with JavaScript

    // get_phash.js
// Using fetch API
const formData = new FormData();
formData.append('file', fileInput.files[0]);

fetch('https://oyyi.xyz/api/image/p-hash', {
  method: 'POST',
  body: formData
})
.then(response => response.json())
.then(hashData => {
  console.log(hashData);
  // Use the hash data for image comparison or identification
})
.catch(error => console.error('Error:', error));
  

Example Response

    // response.json
{
  "hash": "f8e0c0e1f0f8fcfe",
  "hash_type": "phash",
  "hash_size": 16,
  "binary": "1111100011100000110000001110000111110000111110001111110011111110",
  "hex": "f8e0c0e1f0f8fcfe",
  "decimal": 17950536532773458174,
  "normalized": [1, 1, 1, 1, 1, 0, 0, 0, ...]
}
  

How It Works

Perceptual hashing creates a "fingerprint" of an image that remains similar even when the image is modified in non-destructive ways:

1. Preprocessing: The image is converted to grayscale and resized to a small square.
2. Hash Computation: A DCT is applied to identify frequency patterns, then the low-frequency components are used to generate the hash.
3. Hash Formatting: The hash is returned in multiple formats (binary, hex, decimal) for different use cases.

Error Responses

Notes

• Perceptual hashes are designed to be similar for similar images, unlike cryptographic hashes which change completely with even minor modifications.
• To compare two images for similarity, generate a hash for each and calculate the Hamming distance between the hashes (count of differing bits).
• A lower Hamming distance indicates greater similarity. Typically, a distance of 10 or less (out of 64 bits) suggests the images are visually similar.
• The maximum file size allowed is 20MB.
• This operation is not affected by image format, compression, or minor edits.

Use Cases

• Detecting duplicate or near-duplicate images in large collections
• Finding visually similar images across different sources
• Content-based image retrieval systems
• Copyright infringement detection
• Reverse image search functionality
• Tracking modified versions of original images
• Detecting image manipulation or tampering