lab_18.py

#! /usr/bin/python

# autogenerated on 2020-05-28 17:10

# Local Image Descriptors

import cv2
import numpy as np

import os

from utils import *


def sub_histogram(img, size=4):
    height, width, channels = img.shape
    height //= size
    width //= size
    output = np.zeros((size, size, channels, 256), dtype=np.float32)
    for i in range(size):
        for j in range(size):
            for c in range(channels):
                tmp = img[height * i : height * (i + 1), width * j : width * (j + 1), c]
                tmp = get_histogram(tmp, True)
                output[i, j, c] = tmp
    return output


def distance(mat_hist_a, mat_hist_b):
    tmp = 0
    for row_a, row_b in zip(mat_hist_a, mat_hist_b):
        for a, b in zip(row_a, row_b):
            tmp += cv2.compareHist(a, b, cv2.HISTCMP_BHATTACHARYYA)

    return tmp


def search_similar(img, size, path):
    color = len(img.shape) == 3
    img_sub_hist = sub_histogram(img, size)
    min_err = -1
    min_img = None
    for entry in os.scandir(path):
        if entry.is_file:
            name = entry.path
            tmp = load_image(name, color=color)
            tmp_sub_hist = sub_histogram(tmp, size)
            err = distance(img_sub_hist, tmp_sub_hist)
            print(err, name)
            if min_err > err or min_err == -1:
                min_err = err
                min_img = name

    return min_img


def main():
    img = load_image_from_arg(color=True)
    similar = search_similar(img, size=5, path="images")
    print(similar)
    similar = load_image(similar, color=True)
    show_image(("given", img), ("similar", similar), wait=10)


if __name__ == "__main__":
    main()