# 1:read image

    # 3: cluster, I thought: give every pixel (that in orignal image)#    a label , so the label have same shape as image(gray)cls = KMeans(n_clusters=k_cluster).fit_predict(data)cls = cls.reshape(image.shape[0], image.shape[1])

    # 4: create a image containercontainer = np.zeros(shape=(image.shape[0], image.shape[1]))

    # 5: use cluster labels as "gray value" #    and  fill it into aimage containerfor i in range(image.shape[0]):for j in range(image.shape[1]):# cls[i, j]*30 ,because label value is 0, 1, 2# the bright difference betwwen labels is to smallcontainer[i, j] = cls[i, j]*60

    # 6: saver the cluster imagecontainer = container.astype(np.uint8) imageio.imsave(save_name, container)

import numpy as np
import imageio
from sklearn.cluster import KMeansdef image_cluster(image_name, save_name, k_cluster=3):"""cluster by KMeans for RGB image"""# 1:read imageimage = imageio.imread(image_name)# 2: convert (w, h, 3) into (w*h, 3)# R,G and B combine as 3 features # data will be a 2D matrix, each row have 3 values(R/G/B),# and each column has width*height values# this operation convert 3D to 2D, like reshape image2matrix = []for i in range(image.shape[0]):for j in range(image.shape[1]):r_v, g_v, b_v = image[i, j]image2matrix.append([r_v/255.0, g_v/255.0, b_v/255.0])data = np.mat(image2matrix)# 3: cluster, I thought: give every pixel (that in orignal image)#    a label , so the label have same shape as image(gray)cls = KMeans(n_clusters=k_cluster).fit_predict(data)cls = cls.reshape(image.shape[0], image.shape[1])# 4: create a image containercontainer = np.zeros(shape=(image.shape[0], image.shape[1]))# 5: use cluster labels as "gray value" #    and  fill it into aimage containerfor i in range(image.shape[0]):for j in range(image.shape[1]):# cls[i, j]*30 ,because label value is 0, 1, 2# the bright difference betwwen labels is to smallcontainer[i, j] = cls[i, j]*60# 6: saver the cluster imagecontainer = container.astype(np.uint8) imageio.imsave(save_name, container)return Trueimage_cluster("data/vivian.jpg", "results/cluster.jpg")