import numpy as np

import matplotlib.pyplot as plt

# Partie A - Importer une image comme un tableau

import imageio.v3 as imageio

f = imageio.imread('image_avant.png')

# img02 : flou

# img05 : piqué

# img04 : contour

# img10 : emboss

# img09 : verticale

# Partie B - Motif de convolution

# Identité

g = np.array([[0,0,0],[0,1,0],[0,0,0]]) # identité

# Flou

# g = 1/9*np.array([[1,1,1],[1,1,1],[1,1,1]]) # flou 3x3

# g = 1/25*np.ones((5,5)) # flou 5x5

# g = 1/16*np.array([[1,2,1],[2,4,2],[1,2,1]]) # flou gaussien 3x3

# g = 1/256*np.array([[1,2,4,6,41],[4,16,24,16,4],[6,24,36,24,6],[4,16,24,16,4],[1,2,4,6,41]]) # flou gaussien 5x5

# Piqué

# g = np.array([[0,-1,0],[-1,5,-1],[0,-1,0]]) # piqué

# Contour

# g = 1/9*np.array([[1,0,-1],[0,0,0],[-1,0,1]]) # contour

# g = np.array([[0,1,0],[1,-4,1],[0,1,0]]) # contour

# g = np.array([[-1,-1,-1],[-1,8,-1],[-1,-1,-1]]) # contour

g = np.array([[-2,-1,0],[-1,1,1],[0,1,2]]) # 'emboss'

# g = np.array([[-1,2,-1],[-1,2,-1],[-1,2,-1]]) # verticale

# g = np.array([[-1,2,-1],[0,0,0],[-1,2,-1]]) # verticale (Sobel)

# g = np.array([[-1,-1,-1],[2,2,2],[-1,-1,-1]]) # horizontale

# g = np.array([[-1,-1,2],[-1,2,-1],[2,-1,-1]]) # 45 degrés

# Partie C - Calcul de la convolution

from scipy import signal

h = signal.convolve2d(f, g, boundary='fill', mode='same')

# Partie D - Affichage des images avant/après

fig = plt.figure(figsize = (10,5))

ax = plt.subplot(1,2,1)

ax.set_title("Image originale")

# fig, ax = subplots(figsize=(18, 2))

ax.imshow(f, cmap='gray')

# plt.xlim(0, 200)

# plt.ylim(0, 200)

ax = plt.subplot(1,2,2)

ax.set_title("Image sortie")

plt.imshow(h, cmap='gray')

plt.tight_layout()

plt.show()

# Partie E - Sauvegarde de l'image

h = np.clip(h,0,255) # limite les valeurs entre 0 et 255

h = h.astype(np.uint8) # conversion en entier

imageio.imwrite('image_apres.png', h)