from sklearn import datasets
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

# 加载图片
china = datasets.load_sample_image("china.jpg")
plt.imshow(china)

# 把像素点随机打乱
from sklearn.utils import shuffle
img_array_shuffle = shuffle(img_array,random_state=0)

plt.imshow(img_array_shuffle.reshape((h,w,d)))

from sklearn.cluster import KMeans
# 用64个聚类来划分这些像素点
km = KMeans(n_clusters=64,random_state=0)

# 仅用前1000个像素点进行模型训练
km.fit(img_array_shuffle[:1000])

KMeans(algorithm='auto', copy_x=True, init='k-means++', max_iter=300,
    n_clusters=64, n_init=10, n_jobs=1, precompute_distances='auto',
    random_state=0, tol=0.0001, verbose=0)

# 表示这1000个样本对应的聚类标签，展示前100个标签编号
km.labels_[:100]

array([22, 36,  8, 56, 38, 42, 22, 31,  7,  2, 55, 31, 62, 21, 52, 43, 31,
       18,  2,  4, 31, 56, 21,  2, 15,  5,  6, 49, 57, 13,  5, 21, 21,  3,
       21, 47, 21,  2, 47, 32,  5, 42,  5, 33, 45, 56,  5, 57,  2, 38, 47,
        6, 50, 50, 27, 62, 56, 57, 30, 28,  6,  5, 26, 24, 58, 44,  8, 21,
       58, 60, 10, 56, 31, 10, 41,  5, 62, 41, 22,  6, 38, 25, 57, 36, 28,
       21, 49,  2, 21, 48, 14, 15, 57, 22, 47, 63, 47,  2, 41, 34])

# # 用聚类中心点的颜色来代表这个聚类的颜色,展示前10个聚类中心
km.cluster_centers_[:10] 

array([[0.62570806, 0.60261438, 0.53028322],
       [0.15546218, 0.1557423 , 0.12829132],
       [0.82063983, 0.89896801, 0.98462332],
       [0.42039216, 0.43843137, 0.2227451 ],
       [0.69527105, 0.74994233, 0.76516724],
       [0.92174422, 0.9556336 , 0.99514194],
       [0.07058824, 0.0754637 , 0.0508744 ],
       [0.28205128, 0.26395173, 0.19638009],
       [0.46509804, 0.43372549, 0.36901961],
       [0.71328976, 0.41960784, 0.31851852]])

def recreate_img(centers,labels,h,w):
    # 行数的作用是将每个像素点的值，用对应编号的聚类中心代替
    # 按照h和w创建一个空白图片
    img = np.zeros((h,w,3))
    # 通过for循环，遍历img中每一个点，并且从labels中取出下标对应的聚类重新给img赋值
    label_index = 0
    for i in range(h):
        for j in range(w):
            img[i][j] = centers[labels[label_index]]
            label_index += 1
    return img

plt.figure(figsize=(12,6))
axes1 = plt.subplot(121)
axes1.imshow(china)
axes1.set_title("Instaces")

axes2 = plt.subplot(122)
axes2.imshow(img_re)
axes2.set_title("64 colors,K-Means")

Text(0.5,1,'64 colors,K-Means')