[OpenCV] MeanShift Clustering C++ Code

 

#include <iostream>
#include <vector>
#include <ctime>

#include "opencv342.h"

using namespace std;

cv::Mat MeanShift_Cluster(cv::Mat);

static void floodFillPostprocess(cv::Mat& img, const cv::Scalar& colorDiff = cv::Scalar::all(1)) {
  CV_Assert(!img.empty());
  cv::RNG rng = cv::theRNG();
  cv::Mat mask(img.rows + 2, img.cols + 2, CV_8UC1, cv::Scalar::all(0));

  /////////////////////////
  // 추가한 부분
  cv::Mat _mask = mask.clone(), diff = mask.clone();
  cv::Mat cpy = img.clone();
  float val1, val2, val3;
  int nPoints, iTemp, i, j;
  ////////////////////////

  for (int y = 0; y < img.rows; y++) {
    for (int x = 0; x < img.cols; x++) {
      if (mask.at<uchar>(y + 1, x + 1) == 0) {
        mask.copyTo(_mask); // 추가한 부분
        cv::Scalar newVal(rng(256), rng(256), rng(256));
        // img 대신 cpy 로 변경
        cv::floodFill(cpy, mask, cv::Point(x, y), newVal, 0, colorDiff, colorDiff);
        diff = mask - _mask; // 추가한 부분

        /////////////////////////////
        // 추가한 부분
        nPoints = 0;
        val1 = val2 = val3 = 0.0;
        for (i = 0; i < img.rows; i++) {
          for (j = 0; j < img.cols; j++) {
            if (diff.at<uchar>(i + 1, j + 1)) {
              val1 += img.at<cv::Vec3b>(i, j)[0];
              val2 += img.at<cv::Vec3b>(i, j)[1];
			  val3 += img.at<cv::Vec3b>(i, j)[2];
			  nPoints++;
			}
		  }
		}
		for (i = 0; i < img.rows; i++) {
		  for (j = 0; j < img.cols; j++) {
			if (diff.at<uchar>(i + 1, j + 1)) {
			  iTemp = cvRound(val1 / (float)nPoints);
			  iTemp = iTemp > 255 ? 255 : iTemp < 0 ? 0 : iTemp;
			  img.at<cv::Vec3b>(i, j)[0] = (uchar)iTemp;

			  iTemp = cvRound(val2 / (float)nPoints);
			  iTemp = iTemp > 255 ? 255 : iTemp < 0 ? 0 : iTemp;
			  img.at<cv::Vec3b>(i, j)[1] = (uchar)iTemp;

			  iTemp = cvRound(val3 / (float)nPoints);
			  iTemp = iTemp > 255 ? 255 : iTemp < 0 ? 0 : iTemp;
			  img.at<cv::Vec3b>(i, j)[2] = (uchar)iTemp;
			}
		  }
		}
	  }
	}
  }
}

int main() {
  string CLUSTER[5] = { "12003.jpg", "25098.jpg", "97017.jpg", "151087.jpg", "181018.jpg" };
  string path = "testImage/";

  // Mean Shift Clustering
  cout << "Start MeanShift Clustering." << endl;
  for (int i = 0; i < 5; i++) {
    cv::Mat img = cv::imread(path + CLUSTER[i]);
    cv::Mat result;
    result = MeanShift_Cluster(img);
    cv::imwrite(path + "Result_MeanShift/" + CLUSTER[i], result);
  }
  cout << "End MeanShift Clustering." << endl;
}

cv::Mat MeanShift_Cluster(cv::Mat src) {
  cv::Mat result;

  double sp, sr;

  sp = 10; // the spatial window radius
  sr = 10; // the color window radius
  cv::pyrMeanShiftFiltering(src, result, sp, sr);
  floodFillPostprocess(result, cv::Scalar::all(2));

  return result;
}