100 lines
3.7 KiB
Python
100 lines
3.7 KiB
Python
|
from PIL import Image, ImageOps
|
||
|
|
import numpy as np
|
||
|
|
import random as rd
|
||
|
|
import math
|
||
|
|
|
||
|
|
# Lecture de l'image et changement du contraste
|
||
|
|
image = Image.open("./images/2.jpg",mode = 'r')
|
||
|
|
image = ImageOps.autocontrast(image, cutoff=10)
|
||
|
|
|
||
|
|
im = np.asarray(image)
|
||
|
|
|
||
|
|
nb_lignes, nb_colones, nb_subpixel = np.shape(im) # nb_subpixel est le nombre d'élements d'un points (R, G, B) = 3
|
||
|
|
classe = [[-1 for i in range(nb_colones)] for j in range(nb_lignes)] # Tableau des classes de chaque éléments
|
||
|
|
nb_classe = 8
|
||
|
|
|
||
|
|
# initialisation centres
|
||
|
|
centres = []
|
||
|
|
|
||
|
|
for i in range(nb_classe) :
|
||
|
|
ir = rd.randint(0,nb_lignes-1)
|
||
|
|
jr = rd.randint(0,nb_colones-1)
|
||
|
|
while [im[ir,jr,s] for s in range(nb_subpixel)] in centres :
|
||
|
|
ir = rd.randint(0,nb_lignes-1)
|
||
|
|
jr = rd.randint(0,nb_colones-1)
|
||
|
|
centres.append([im[ir,jr,s] for s in range(nb_subpixel)]) # On prend des pixels aléatoires dans l'image, et on vérifie qu'ils sont tous differents
|
||
|
|
|
||
|
|
# distance (voire pour prendre plus efficace ?)
|
||
|
|
|
||
|
|
def distance(e1: list, e2: list) : # distance entre deux couleurs
|
||
|
|
d = len(e1)
|
||
|
|
res = 0
|
||
|
|
for i in range(d) :
|
||
|
|
res = max(res,(e1[i]-e2[i])**2)
|
||
|
|
return(res)
|
||
|
|
|
||
|
|
# ranger au plus proche des centres
|
||
|
|
def classer() : # permet de classer les elements dans chaque classe
|
||
|
|
global classe, im, nb_lignes, nb_colones, centres
|
||
|
|
b = False
|
||
|
|
for i in range(nb_lignes) : # On parcours toute l'image
|
||
|
|
for j in range(nb_colones) : # --
|
||
|
|
dmin = 0
|
||
|
|
if classe[i][j] == -1 : # si il n'y a pas encore de classe
|
||
|
|
dmin = math.inf # la distance est la minimum
|
||
|
|
else :
|
||
|
|
dmin = distance(im[i,j], centres[classe[i][j]])
|
||
|
|
for m in range(nb_classe) :
|
||
|
|
dist = distance(list(im[i,j]), centres[m])
|
||
|
|
if dist<dmin :
|
||
|
|
dmin = dist
|
||
|
|
classe[i][j] = m
|
||
|
|
b = True
|
||
|
|
return(b)
|
||
|
|
|
||
|
|
|
||
|
|
# les centres bougent
|
||
|
|
def barycentre() :
|
||
|
|
global classe, im, nb_lignes, nb_colones, centres
|
||
|
|
dico = dict()
|
||
|
|
for i in range(nb_classe) :
|
||
|
|
dico[i] = []
|
||
|
|
for i in range(nb_lignes) :
|
||
|
|
for j in range(nb_colones) :
|
||
|
|
dico[classe[i][j]].append(im[i,j])
|
||
|
|
for i in range(nb_classe) :
|
||
|
|
r, g, b = 0, 0, 0
|
||
|
|
for e in dico[i] :
|
||
|
|
r += e[0]
|
||
|
|
g += e[1]
|
||
|
|
b += e[2]
|
||
|
|
m = len(dico[i])
|
||
|
|
centres[i] = [r//m, g//m, b//m]
|
||
|
|
# kmoyennes
|
||
|
|
i=0
|
||
|
|
while classer() :
|
||
|
|
i+=1
|
||
|
|
print("\nstep"+str(i))
|
||
|
|
print(centres)
|
||
|
|
barycentre()
|
||
|
|
im_classes = np.array([[centres[classe[i][j]] for j in range(nb_colones)] for i in range(nb_lignes)])
|
||
|
|
res = Image.fromarray(im_classes, mode='RGB')
|
||
|
|
|
||
|
|
|
||
|
|
# affichage,sauvegarde
|
||
|
|
|
||
|
|
colors = [[0,0,0],[150,0,0],[0,150,0],[0,0,150],[128,128,0],[0,160,160],[192,0,192],[230,230,230]] # liste des couleurs pour bien différencier les classes
|
||
|
|
|
||
|
|
# Enregistrer toutes les images
|
||
|
|
for lettre in range(nb_classe):
|
||
|
|
im_classes = np.array([[im[i][j] if classe[i][j] == lettre else (0, 0, 0) for j in range(nb_colones)] for i in range(nb_lignes)],dtype=np.uint8)
|
||
|
|
print(im_classes)
|
||
|
|
res = Image.fromarray(im_classes, mode="RGB")
|
||
|
|
res.show()
|
||
|
|
res.save(f"./out/2_{lettre}.jpg")
|
||
|
|
|
||
|
|
|
||
|
|
b = np.array([[colors[classe[i][j]] for j in range(nb_colones)] for i in range(nb_lignes)],dtype=np.uint8)
|
||
|
|
res = Image.fromarray( b,mode="RGB")
|
||
|
|
res.show()
|
||
|
|
res.save("./out/out.jpg")
|