phash.py

phash.py

import numpy as np
from scipy.fftpack import dct


def hash_algo(pil_img, size=10):
    """
    Get perceptual hash of the input image.

    Args:
        image_array: numpy array that corresponds to the image.

    Returns:
        A string representing the perceptual hash of the image.
    """
    image_array = np.array(pil_img)
    dct_coef = dct(dct(image_array, axis=0), axis=1)
    dct_reduced_coef = dct_coef[:size, :size]
    median_coef_val = np.median(np.ndarray.flatten(dct_reduced_coef)[1:])
    hash_mat = dct_reduced_coef >= median_coef_val
    return hash_mat



def str_image_hash(pil_image):
    hash_str = hash_algo(np.array(pil_image))
    return ''.join([str(v) for v in hash_str.astype('int').reshape(-1)])


def bin_to_hex_hash_sim(bin_hash_str):
    n = 4
    sub_strings = [format(int(bin_hash_str[i:i+n], 2), 'x') for i in range(0, len(bin_hash_str), n)]

    return "".join(sub_strings)


def hex_to_bin_hash_sim(hex_hash_str):
    n = 1
    sub_strings = [format(int(hex_hash_str[i:i+n], 16), 'b').zfill(4) for i in range(0, len(hex_hash_str), n)]

    return "".join(sub_strings)