DCT test

dct_test.py

import numpy as np
from scipy.fftpack import dct, idct
import matplotlib.pyplot as plt
from PIL import Image
import matplotlib.image

def dct_2d(x):
    return dct(dct(x, axis=0, norm='ortho'), axis=1, norm='ortho')

def idct_2d(x):
    return idct(idct(x, axis=0 , norm='ortho'), axis=1 , norm='ortho')

def compress(image, n):
    w = np.zeros_like(image)
    w[0:n, 0:n, :] = 1
    k = dct_2d(image) * w
    return idct_2d(k)

def discrete(image):
    r0 = (image[:, :, 0] > image[:, :, 1]).astype(float) * (image[:, :, 0] > image[:, :, 2]).astype(float)
    r1 = (image[:, :, 1] > image[:, :, 0]).astype(float) * (image[:, :, 1] > image[:, :, 2]).astype(float)
    r2 = (image[:, :, 2] > image[:, :, 0]).astype(float) * (image[:, :, 2] > image[:, :, 1]).astype(float)
    return np.dstack([r0, r1, r2])

  
image = matplotlib.image.imread("1.png")

for i in range(1, 200):
    result = discrete(compress(image, i))
    matplotlib.image.imsave(f"v/{i:03}.png", result)