颜色量化是减少图像中不同颜色数量的过程。通常，目的是尽可能保留图像的颜色外观，同时减少颜色数量，无论是用于内存限制还是压缩。实际上，著名的QBIC CBIR系统（图像搜索引擎的原始CBIR系统之一）可以利用二次距离中的量化颜色直方图来计算相似度。
OpenCv指定聚类颜色和聚类数量，聚类算法有很多种（几十种），K-Means是聚类算法中的最常用的一种，算法最大的特点是简单，好理解，运算速度快，但是只能应用于连续型的数据，并且一定要在聚类前需要手工指定要分成几类。
OpenCvSharp代码如下：
Mat srcImage = Cv2.ImRead(@"C:\Users\Desktop\123.png"); //五个颜色，聚类之后的颜色随机从这里面选择 Scalar[] colorTab = { new Scalar(0,0,255), new Scalar(0,255,0), new Scalar(255,0,0), new Scalar(0,255,255), new Scalar(255,0,255) }; int width = srcImage.Cols;//图像的宽 int height = srcImage.Rows;//图像的高 int channels = srcImage.Channels();//图像的通道数 //初始化一些定义 int sampleCount = width * height;//所有的像素 int clusterCount = 4;//分类数 Mat points = new Mat(sampleCount, channels, MatType.CV_32F, Scalar.All(10));//points用来保存所有的数据 Mat labels = new Mat();//聚类后的标签 Mat center = new Mat(clusterCount, 1, points.Type());//聚类后的类别的中心 //将图像的RGB像素转到到样本数据 int index; for (int i = 0; i < srcImage.Rows; i++) { for (int j = 0; j < srcImage.Cols; j++) { index = i * width + j; Vec3b bgr = srcImage.At<Vec3b>(i, j); //将图像中的每个通道的数据分别赋值给points的值 points.Set<float>(index, 0, bgr.Item0); points.Set<float>(index, 1, bgr.Item1); points.Set<float>(index, 2, bgr.Item2); } } //运行K-means算法 //MAX_ITER也可以称为COUNT最大迭代次数，EPS最高精度,10表示最大的迭代次数，0.1表示结果的精确度 TermCriteria criteria = new TermCriteria( CriteriaType.Eps & CriteriaType.Count, 10, 0.1); Cv2.Kmeans(points, clusterCount, labels, criteria, 3, KMeansFlags.PpCenters, center); //显示图像分割结果 Mat result = Mat.Zeros(srcImage.Size(), srcImage.Type());//创建一张结果图 for (int i = 0; i < srcImage.Rows; i++) { for (int j = 0; j < srcImage.Cols; j++) { index = i * width + j; int label = labels.At<int>(index);//每一个像素属于哪个标签 Vec3b vec3B = new Vec3b(); vec3B.Item0 = (byte)colorTab[label][0];//对结果图中的每一个通道进行赋值 vec3B.Item1 = (byte)colorTab[label][1]; vec3B.Item2 = (byte)colorTab[label][2]; result.Set<Vec3b>(i, j, vec3B); } } Cv2.ImShow("Kmeans", result);原图 聚类结果
Python K-Means聚类算法进行颜色量化
# KMeans聚类算法图像颜色量化 # import the necessary packages from sklearn.cluster import MiniBatchKMeans import numpy as np import argparse import cv2 # load the image and grab its width and height image = cv2.imread("C:/Users/Desktop/123.png") (h, w) = image.shape[:2] # convert the image from the RGB color space to the L*a*b* # color space -- since we will be clustering using k-means # which is based on the euclidean distance, we'll use the # L*a*b* color space where the euclidean distance implies # perceptual meaning image = cv2.cvtColor(image, cv2.COLOR_BGR2LAB) # reshape the image into a feature vector so that k-means # can be applied image = image.reshape((image.shape[0] * image.shape[1], 3)) # apply k-means using the specified number of clusters and # then create the quantized image based on the predictions clt = MiniBatchKMeans(n_clusters = 4) labels = clt.fit_predict(image) quant = clt.cluster_centers_.astype("uint8")[labels] # reshape the feature vectors to images quant = quant.reshape((h, w, 3)) image = image.reshape((h, w, 3)) # convert from L*a*b* to RGB quant = cv2.cvtColor(quant, cv2.COLOR_LAB2BGR) image = cv2.cvtColor(image, cv2.COLOR_LAB2BGR) # display the images and wait for a keypress cv2.imshow("image", np.hstack([image, quant])) cv2.waitKey(0)左原图 右效果图 参数K=4
左原图 右效果图 参数K=6 左原图 右效果图 参数K=12
