【OpenCV图像处理】1.33 基于距离变换与分水岭的图像分割

1. 相关理论

• 什么是图像分割(Image Segmentation)

  • 图像分割(Image Segmentation)是图像处理最重要的处理手段之一
  • 图像分割的目标是将图像中像素根据一定的规则分为若干(N)个cluster集合，每个集合包含一类像素。
  • 根据算法分为监督学习方法和无监督学习方法，图像分割的算法多数都是无监督学习方法 - KMeans

• 距离变换与分水岭介绍

  • 距离变换常见算法有两种
    • 不断膨胀/ 腐蚀得到
    • 基于倒角距离
  • 分水岭变换常见的算法
    • 基于浸泡理论实现

• 相关API

  cv::distanceTransform(
      InputArray  src, OutputArray dst,  
      OutputArray  labels,  int  distanceType, 
      int maskSize,  int labelType=DIST_LABEL_CCOMP)
  
  distanceType = DIST_L1/DIST_L2,
  maskSize = 3x3,最新的支持5x5，推荐3x3、
  labels 离散维诺图输出
  dst 输出8位或者32位的浮点数，单一通道，大小与输入图像一致
  
  cv::watershed(InputArray image, InputOutputArray  markers)
  

2. 代码 & 效果展示

• 处理流程

  1. 将白色背景变成黑色-目的是为后面的变换做准备
  2. 使用filter2D与拉普拉斯算子实现图像对比度提高，sharp
  3. 转为二值图像通过threshold
  4. 距离变换
  5. 对距离变换结果进行归一化到[0~1]之间
  6. 使用阈值，再次二值化，得到标记
  7. 腐蚀得到每个Peak - erode
  8. 发现轮廓 – findContours
  9. 绘制轮廓- drawContours
  10. 分水岭变换 watershed
  11. 对每个分割区域着色输出结果

• 代码：

  #include <iostream>
  #include <opencv2/opencv.hpp>
  #include <opencv2/imgproc/types_c.h>
  
  using namespace std;
  using namespace cv;
  
  #ifndef P33
  #define P33 33
  #endif
  
  int main() {
      std::string path = "../circle.JPG";
      cv::Mat img = cv::imread(path, 5);
  
      string str_input = "input image";
      string str_output = "output image";
  
      if (img.empty()) {
          std::cout << "open file failed" << std::endl;
          return -1;
      }
  
      namedWindow(str_input, WINDOW_AUTOSIZE);
      namedWindow(str_output, WINDOW_AUTOSIZE);
      imshow(str_input, img);
      
  #if P34
      Mat src = img;
  
      // 1. change background
      for (int row = 0; row < src.rows; row++) {
          for (int col = 0; col < src.cols; col++) {
              if (src.at<Vec3b>(row, col) == Vec3b(255, 255, 255)) {
                  src.at<Vec3b>(row, col)[0] = 0;
                  src.at<Vec3b>(row, col)[1] = 0;
                  src.at<Vec3b>(row, col)[2] = 0;
              }
          }
      }
      namedWindow("black background", WINDOW_AUTOSIZE);
      imshow("black background", src);
  
      // sharpen
      Mat kernel = (Mat_<float>(3, 3) << 1, 1, 1, 1, -8, 1, 1, 1, 1);
      Mat imgLaplance;
      Mat sharpenImg = src;
      filter2D(src, imgLaplance, CV_32F, kernel, Point(-1, -1), 0, BORDER_DEFAULT);
      src.convertTo(sharpenImg, CV_32F);
      Mat resultImg = sharpenImg - imgLaplance;
  
      resultImg.convertTo(resultImg, CV_8UC3);
      imgLaplance.convertTo(imgLaplance, CV_8UC3);
      imshow("sharpen image", resultImg);
      // src = resultImg; // copy back
  
      // convert to binary
      Mat binaryImg;
      cvtColor(src, resultImg, CV_BGR2GRAY);
      threshold(resultImg, binaryImg, 40, 255, THRESH_BINARY | THRESH_OTSU);
      imshow("binary image", binaryImg);
  
      Mat distImg;
      distanceTransform(binaryImg, distImg, DIST_L1, 3, 5);
      normalize(distImg, distImg, 0, 1, NORM_MINMAX);
      imshow("distance result", distImg);
  
      // binary again
      threshold(distImg, distImg, .4, 1, THRESH_BINARY);
      Mat k1 = Mat::ones(13, 13, CV_8UC1);
      erode(distImg, distImg, k1, Point(-1, -1));
      imshow("distance binary image", distImg);
  
      // markers
      Mat dist_8u;
      distImg.convertTo(dist_8u, CV_8U);
      vector<vector<Point>> contours;
      findContours(dist_8u, contours, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE, Point(0, 0));
  
      // create makers
      Mat markers = Mat::zeros(src.size(), CV_32SC1);
      for (size_t i = 0; i < contours.size(); i++) {
          drawContours(markers, contours, static_cast<int>(i), Scalar::all(static_cast<int>(i) + 1), -1);
      }
      circle(markers, Point(5, 5), 3, Scalar(255, 255, 255), -1);
      //imshow("my markers", markers);
      //imshow("my markers", markers*1000);
  
      // perform watershed
      watershed(src, markers);
      Mat mark = Mat::zeros(markers.size(), CV_8UC1);
      markers.convertTo(mark, CV_8UC1);
      bitwise_not(mark, mark, Mat());
      imshow("watershed image", mark);
  
      // generate random color
      vector<Vec3b> colors;
      for (size_t i = 0; i < contours.size(); i++) {
          int r = theRNG().uniform(0, 255);
          int g = theRNG().uniform(0, 255);
          int b = theRNG().uniform(0, 255);
          colors.push_back(Vec3b((uchar)b, (uchar)g, (uchar)r));
      }
  
      // fill with color and display final result
      Mat dst = Mat::zeros(markers.size(), CV_8UC3);
      for (int row = 0; row < markers.rows; row++) {
          for (int col = 0; col < markers.cols; col++) {
              int index = markers.at<int>(row, col);
              if (index > 0 && index <= static_cast<int>(contours.size())) {
                  dst.at<Vec3b>(row, col) = colors[index - 1];
              }
              else {
                  dst.at<Vec3b>(row, col) = Vec3b(0, 0, 0);
              }
          }
      }
      imshow("Final Result", dst);
  #endif
  
      cv::waitKey(0);
      cv::destroyAllWindows();
      return 0;
  }	
  

效果展示：
图像比较多，只展示原图和经过分水岭处理后的图形：