Introduction

This tutorial will show you how to extract the contours from a binary image. The contour extraction algorithm is based on [38] article and most of the implementation has been ported from [16] library.

The function to call is vp::findContours(const vpImage<unsigned char> &, vpContour &, std::vector<std::vector<vpImagePoint> > &, const vpContourRetrievalType&):

the first argument is the image where '0' pixel value means the background and '1' pixel value means the foreground. Other values are not allowed.
the second argument is a vp::vpContour structure that contains the list of contours in a tree
the third argument is the list of contours
the last argument is an option to choose the type of contour extraction, see vp::vpContourRetrievalType

The vp::vpContour structure is composed of:

std::vector< vp::vpContour * > m_children, the list of children contours for the current contour
vp::vpContourType m_contourType, the type of contour (vp::CONTOUR_OUTER or vp::CONTOUR_HOLE)
vp::vpContour * m_parent, the parent contour for the current contour
std::vector< vpImagePoint > m_points, the list of contour points
the first or top level contour is called the root contour (with vp::CONTOUR_HOLE type by default) and contains in m_children the list of contours

The different contour extraction methods are:

vp::CONTOUR_RETR_TREE, all the contours are extracted and stored in a hierarchical tree.
vp::CONTOUR_RETR_LIST, all the contours are extracted and stored in a list. The top level contour contains in m_children the list of all the extracted contours.
vp::CONTOUR_RETR_EXTERNAL, only the external contours are extracted and stored in a list. The top level contour contains in m_children the list of the external extracted contours.

The next section will provide a concrete example for better understanding.

Example code

The following example also available in tutorial-contour.cpp will demonstrate on a sample image the result of each of these methods:

#include <cstdlib>
#include <iostream>
#include <visp3/core/vpImage.h>
#include <visp3/gui/vpDisplayGDI.h>
#include <visp3/gui/vpDisplayOpenCV.h>
#include <visp3/gui/vpDisplayX.h>
#include <visp3/io/vpImageIo.h>
#if defined(VISP_HAVE_MODULE_IMGPROC) && (defined(VISP_HAVE_X11) || defined(VISP_HAVE_GDI) || defined(VISP_HAVE_OPENCV))
#include <visp3/imgproc/vpImgproc.h>
namespace
{
void displayContourInfo(const vp::vpContour &contour, int level)
{
  std::cout << "\nContour:" << std::endl;
  std::cout << "\tlevel: " << level << std::endl;
  std::cout << "\tcontour type: " << (contour.m_contourType == vp::CONTOUR_OUTER ? "outer contour" : "hole contour")
            << std::endl;
  std::cout << "\tcontour size: " << contour.m_points.size() << std::endl;
  std::cout << "\tnb children: " << contour.m_children.size() << std::endl;
  for (std::vector<vp::vpContour *>::const_iterator it = contour.m_children.begin(); it != contour.m_children.end();
       ++it) {
    displayContourInfo(**it, level + 1);
  }
}
void drawContoursTree(vpImage<vpRGBa> &I, const vp::vpContour &contour)
{
  std::vector<std::vector<vpImagePoint> > contours;
  contours.push_back(contour.m_points);
  vp::drawContours(I, contours, contour.m_contourType == vp::CONTOUR_OUTER ? vpColor::red : vpColor::green);
  for (std::vector<vp::vpContour *>::const_iterator it = contour.m_children.begin(); it != contour.m_children.end();
       ++it) {
    drawContoursTree(I, **it);
  }
}
} // namespace
#endif
int main(int argc, const char **argv)
{
#if defined(VISP_HAVE_MODULE_IMGPROC) && (defined(VISP_HAVE_X11) || defined(VISP_HAVE_GDI) || defined(VISP_HAVE_OPENCV))
  std::string input_filename = "grid36-03.pgm";
  bool white_foreground = false;
  vp::vpContourRetrievalType extraction_method = vp::CONTOUR_RETR_TREE;
  for (int i = 1; i < argc; i++) {
    if (std::string(argv[i]) == "--input" && i + 1 < argc) {
      input_filename = std::string(argv[i + 1]);
    } else if (std::string(argv[i]) == "--white_foreground") {
      white_foreground = true;
    } else if (std::string(argv[i]) == "--method" && i + 1 < argc) {
      extraction_method = (vp::vpContourRetrievalType)atoi(argv[i + 1]);
    } else if (std::string(argv[i]) == "--help" || std::string(argv[i]) == "-h") {
      std::cout << "Usage: " << argv[0]
                << " [--input <input image>] [--method <0: "
                   "CONTOUR_RETR_TREE, 1: CONTOUR_RETR_LIST, 2: "
                   "CONTOUR_RETR_EXTERNAL>]"
                   " [--white_foreground] [--help]"
                << std::endl;
      return EXIT_SUCCESS;
    }
  }
  vpImage<unsigned char> I;
  vpImageIo::read(I, input_filename);
  vpImage<unsigned char> I_bin(I.getHeight(), I.getWidth());
  vpImage<vpRGBa> I_draw_contours(I.getHeight(), I.getWidth());
#ifdef VISP_HAVE_X11
  vpDisplayX d, d2;
#elif defined(VISP_HAVE_GDI)
  vpDisplayGDI d, d2;
#elif defined(VISP_HAVE_OPENCV)
  vpDisplayOpenCV d, d2;
#endif
  d.init(I_bin, 0, 0, "After binarisation");
  d2.init(I_draw_contours, I_bin.getWidth(), 10, "Contours");
  vp::autoThreshold(I, vp::AUTO_THRESHOLD_OTSU, white_foreground ? 0 : 1, white_foreground ? 1 : 0);
  for (unsigned int i = 0; i < I_bin.getSize(); i++) {
    I_bin.bitmap[i] = 255 * I.bitmap[i];
  }
  vp::vpContour vp_contours;
  std::vector<std::vector<vpImagePoint> > contours;
  vp::findContours(I, vp_contours, contours, extraction_method);
  vp::drawContours(I_draw_contours, contours, vpColor::red);
  vpDisplay::display(I_bin);
  vpDisplay::display(I_draw_contours);
  vpDisplay::displayText(I_draw_contours, 20, 20, "Click to draw outer / hole contours.", vpColor::red);
  vpDisplay::flush(I_bin);
  vpDisplay::flush(I_draw_contours);
  vpDisplay::getClick(I_draw_contours);
  I_draw_contours = 0;
  drawContoursTree(I_draw_contours, vp_contours);
  displayContourInfo(vp_contours, 0);
  vpDisplay::display(I_bin);
  vpDisplay::display(I_draw_contours);
  vpDisplay::displayText(I_draw_contours, 20, 20, "Click to quit.", vpColor::red);
  vpDisplay::displayText(I_draw_contours, 20, I_draw_contours.getWidth() - 200, "Outer contour", vpColor::red);
  vpDisplay::displayText(I_draw_contours, 20, I_draw_contours.getWidth() - 100, "Hole contour", vpColor::green);
  vpDisplay::flush(I_bin);
  vpDisplay::flush(I_draw_contours);
  vpDisplay::getClick(I_draw_contours);
  return EXIT_SUCCESS;
#else
  (void)argc;
  (void)argv;
  return 0;
#endif
}

These functions are provided in a vp:: namespace and accessible using this include:

#include <visp3/imgproc/vpImgproc.h>

The first steps are:

read an image in grayscale
vpImage<unsigned char> I;
vpImageIo::read(I, input_filename);
threshold / binarize the image, here with the Otsu method.
vp::autoThreshold(I, vp::AUTO_THRESHOLD_OTSU, white_foreground ? 0 : 1, white_foreground ? 1 : 0);
If the object of interest is in white in the image, the formula for the binarization is:
$I_{bin}\left ( i,j \right ) = \left \{ \begin{matrix} 0 \text{ if } I_{src}\left ( i,j \right ) < \text{threshold} \\ 1 \text{ otherwise} \end{matrix} \right.$
If the object of interest is in black in the image, the formula for the binarization is:
$I_{bin}\left ( i,j \right ) = \left \{ \begin{matrix} 1 \text{ if } I_{src}\left ( i,j \right ) < \text{threshold} \\ 0 \text{ otherwise} \end{matrix} \right.$
extract the contours (by default, it is the vp::CONTOUR_RETR_TREE method)
vp::vpContour vp_contours;
std::vector<std::vector<vpImagePoint> > contours;
vp::findContours(I, vp_contours, contours, extraction_method);
draw the contours if wanted
vp::drawContours(I_draw_contours, contours, vpColor::red);

The result images for each step are:

Input image

Image after binarization using the Otsu method

Contours extracted and displayed on a new image

To understand how the hierarchical contours extraction works, let's switch on another example. In a terminal, run:

$ ./tutorial-contour --input Contours_tree.pgm --white_foreground

The image after binarisation:

Image after binarization using the Otsu method

Instead of drawing all the contours with the same color, we can assign a first color for vp::CONTOUR_OUTER contour and a second color for vp::CONTOUR_HOLE contour.

The function to navigate in the contour tree is the following:

void drawContoursTree(vpImage<vpRGBa> &I, const vp::vpContour &contour)
{
  std::vector<std::vector<vpImagePoint> > contours;
  contours.push_back(contour.m_points);
  vp::drawContours(I, contours, contour.m_contourType == vp::CONTOUR_OUTER ? vpColor::red : vpColor::green);
  for (std::vector<vp::vpContour *>::const_iterator it = contour.m_children.begin(); it != contour.m_children.end();
       ++it) {
    drawContoursTree(I, **it);
  }
}

The call to draw the hierarchical contours:

drawContoursTree(I_draw_contours, vp_contours);

The result image is:

Contours extracted and displayed on a new image, in red outer contours, in green hole contours

To display the hierarchy, we can use this function:

void displayContourInfo(const vp::vpContour &contour, int level)
{
  std::cout << "\nContour:" << std::endl;
  std::cout << "\tlevel: " << level << std::endl;
  std::cout << "\tcontour type: " << (contour.m_contourType == vp::CONTOUR_OUTER ? "outer contour" : "hole contour")
            << std::endl;
  std::cout << "\tcontour size: " << contour.m_points.size() << std::endl;
  std::cout << "\tnb children: " << contour.m_children.size() << std::endl;
  for (std::vector<vp::vpContour *>::const_iterator it = contour.m_children.begin(); it != contour.m_children.end();
       ++it) {
    displayContourInfo(**it, level + 1);
  }
}

For the vp::CONTOUR_RETR_TREE method, the output is:

Contour:
level: 0
contour type: hole contour
contour size: 0
nb children: 3

Contour:
level: 1
contour type: outer contour
contour size: 438
nb children: 0

Contour:
level: 1
contour type: outer contour
contour size: 748
nb children: 0

Contour:
level: 1
contour type: outer contour
contour size: 2012
nb children: 1

Contour:
level: 2
contour type: hole contour
contour size: 1906
nb children: 1

Contour:
level: 3
contour type: outer contour
contour size: 1610
nb children: 1

Contour:
level: 4
contour type: hole contour
contour size: 1494
nb children: 1

Contour:
level: 5
contour type: outer contour
contour size: 792
nb children: 2

Contour:
level: 6
contour type: hole contour
contour size: 372
nb children: 0

Contour:
level: 6
contour type: hole contour
contour size: 392
nb children: 0

The top level contour is always the root contour with zero contour point and which contains the list of contours.

For the vp::CONTOUR_RETR_EXTERNAL method, the output is:

Contour:
level: 0
contour type: hole contour
contour size: 0
nb children: 3
Contour:
level: 1
contour type: outer contour
contour size: 438
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 748
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 2012
nb children: 0

The result image is:

External contours extracted and displayed on a new image

For the vp::CONTOUR_RETR_LIST method, the output is:

Contour:
level: 0
contour type: hole contour
contour size: 0
nb children: 9
Contour:
level: 1
contour type: outer contour
contour size: 438
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 748
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 2012
nb children: 0
Contour:
level: 1
contour type: hole contour
contour size: 1906
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 1610
nb children: 0
Contour:
level: 1
contour type: hole contour
contour size: 1494
nb children: 0
Contour:
level: 1
contour type: outer contour
contour size: 792
nb children: 0
Contour:
level: 1
contour type: hole contour
contour size: 372
nb children: 0
Contour:
level: 1
contour type: hole contour
contour size: 392
nb children: 0

Next tutorial

You can now read the Tutorial: Connected-components labeling, for a similar method to extract the connected-components in a grayscale image.

Table of Contents

Introduction

Example code

Next tutorial