This example was based on the kmeans.cpp algorithm and answers to the proposed exercises at [agostinhobritojr.github.io]https://agostinhobritojr.github.io/tutorial/pdi/#_exerc%C3%ADcios_8).

The following steps were used to accomplish this effect:

1. Choose k as the number of classes for the vectors xi of N samples, i = 1,2, ⋯, N.
2. Choose m1, m2, ⋯, mk as initial approximations for the class centers.
3. Sort each sample xi using, for example, a minimum distance classifier (Euclidean distance).
4. Recalculate the averages mj using the result of the previous step.
5. If the new averages are consistent (do not change considerably), finalize the algorithm. If not, recalculate the centers and redo the classification.

#include <opencv2/opencv.hpp>
#include <cstdlib>
#include <ctime>

using namespace cv;

int main( int argc, char** argv ){
  for(int i = 0; i < 10; i++){

    int nClusters = 20;
    Mat labels;
    int nRounds = 1;
    Mat centers;
    char name[30];

    if(argc!=3){
    exit(0);
    }

    Mat img = imread( argv[1], CV_LOAD_IMAGE_COLOR);
    Mat samples(img.rows * img.cols, 3, CV_32F);

    for( int y = 0; y < img.rows; y++ ){
      for( int x = 0; x < img.cols; x++ ){
        for( int z = 0; z < 3; z++){
          samples.at<float>(y + x*img.rows, z) = img.at<Vec3b>(y,x)[z];
      }
    }
    }

    kmeans(samples,
       nClusters,
       labels,
       TermCriteria(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS, 10000, 0.0001),
       nRounds,
       KMEANS_RANDOM_CENTERS,
       centers );


    Mat labeled( img.size(), img.type() );
    for( int y = 0; y < img.rows; y++ ){
      for( int x = 0; x < img.cols; x++ ){
      int indice = labels.at<int>(y + x*img.rows,0);
      labeled.at<Vec3b>(y,x)[0] = (uchar) centers.at<float>(indice, 0);
      labeled.at<Vec3b>(y,x)[1] = (uchar) centers.at<float>(indice, 1);
      labeled.at<Vec3b>(y,x)[2] = (uchar) centers.at<float>(indice, 2);
    }
    }
    sprintf(name,  "test%d.jpg", i);
    imwrite(name, labeled);
    //waitKey( 0 );
  }
}