如何根据颜色将图像进行聚类实例
第一步：得到图片的RGB模型矩阵
第二步：将RGB模型转化为HSV模型
第三步：将HSV转化为n维的特征向量
第四步：调用K-means对特征矩阵进行聚类
上代码：
import cv2 import numpy as np import os from scipy.cluster.vq import * import shutil import colorsys
class imageKmeans(object):#初始化def __init__(self, k):self.basedir = os.path.dirname(os.path.abspath(__file__)) # 项目根目录self.imagedir = os.path.join(self.basedir, 'flowers') # 图片根目录self.k = k #簇的个数for i in range(self.k): # 创建k个文件夹用于存放分类后的图片clusterDir = os.path.join(self.basedir, 'cluster-{}'.format(i))if not os.path.isdir(clusterDir):os.mkdir(clusterDir)# 返回图片一维数组def _loadImages(self):images = os.listdir(self.imagedir)imagesfiles = [os.path.join(self.imagedir, image) for image in images]return imagesfiles# 将h,s,v合并def _hsvToL(self, h, s, v):OH = 0if (h <= 20 or h > 315):OH = 0if (h > 20 and h <= 40):OH = 1if (h > 40 and h <= 75):OH = 2if (h > 75 and h <= 155):OH = 3if (h > 155 and h <= 190):OH = 4if (h > 190 and h <= 271):OH = 5if (h > 271 and h <= 295):OH = 6if (h > 295 and h <= 315):OH = 7OS = 0if (s >= 0 and s <= 0.2):OS = 0if (s > 0.2 and s < 0.7):OS = 1if (s > 0.7 and s <= 1.0):OS = 2OV = 0if (v >= 0 and v <= 0.2):OV = 0if (v > 0.2 and v <= 0.7):OV = 1if (v > 0.7 and v <= 1.0):OV = 2L = 9 * OH + 3 * OS + OVassert L >= 0 and L <= 71return L# 提取图片特征def _getColor(self,oneImage):image = cv2.imread(oneImage)image = cv2.resize(image, (600, 600))Row = image.shape[0] # 行600Col = image.shape[1] # 列600vector = [0]*12for row in range(Row):for col in range(Col):b, g, r = image[row][col]h, s, v = colorsys.rgb_to_hsv(r / 255., g / 255., b / 255.)h *= 360L = self._hsvToL(h,s,v)vector[L//6] += L #L//6范围 0~11Lsum = sum(vector)result = [v*1.0/Lsum for v in vector]return result#调用Kmeansdef getCluster(self):images = self._loadImages() #图片一维数组(绝对地址)Avector = []for oneImage in images:Vector = self._getColor(oneImage)print(Vector)Avector.append(Vector)image_list = np.reshape(Avector,(len(images),-1))cu, fangcha = kmeans(image_list, self.k)fenlei, juli = vq(image_list, cu)print(fenlei)for i in range(len(images)):for j in range(self.k):if ( fenlei[i] == j ):shutil.copyfile(images[i], os.path.join(os.path.join(self.basedir,'cluster-{}'.format(j)), os.path.basename(images[i])))#主函数
if __name__ == '__main__':imagekmeans = imageKmeans(3)imagekmeans.getCluster()
转自：https://github.com/CekerLee/ImageKmeans