Segment an image using two-dimensional pixel classification.

class_2dim_sup classifies the points in two-channel images using a two-dimensional feature space. For each point, two gray values (one from each image) are used as features. The feature space is represented by the input region. The classification is done as follows:

A point from the input region of an image is accepted if the point (g_{r},g_{c}), which is determined by the respective gray values, is contained in the region FeatureSpace. g_{r} is here a gray value from the image ImageRow, while g_{c} is the corresponding gray value from ImageCol.

Let P be a point with the coordinates P=(R,C), g_{r} be the gray value at position (R,C) in the image ImageRow, and g_{c} be the gray value at position (R,C) in the image ImageCol. Then the point P is aggregated into the output region if

      (g_r,g_c)  is an element of FeatureSpace

For the generation of FeatureSpace, see histo_2dim. The feature space can be modified by applying region transformation operators, such as rank_region, dilation1, shape_trans, elliptic_axis, etc., before calling class_2dim_sup.

The parameters ImageCol and ImageRow must contain an equal number of images with the same size. The image points are taken from the intersection of the domains of both images (see reduce_domain).

#include "HIOStream.h"
#if !defined(USE_IOSTREAM_H)
using namespace std;
#endif
#include "HalconCpp.h"

int main (int argc, char *argv[])
{
  if (argc != 2)
  {
    cout << "Usage : " << argv[0] << " 'image' " << endl;
    return (-1);
  }

  HRegion   feats, cd2reg;
  HImage    image (argv[1]),
            text1, text2,
            mean1, mean2,
            histo;

  HWindow   win;
  long      nc;

  if ((nc = image.CountChannels ()) != 3)
  {
    cout << argv[1] << " is not a rgb-image " << endl;
    return (-2);
  }

  image.Display (win);

  win.SetColor ("green");
  cout << "Draw the region of interrest " << endl;

  HRegion  region = win.DrawRegion ();

  text1 = image.TextureLaws ("el", 2, 5);  
  mean1 = text1.MeanImage (21, 21);
  text2 = mean1.TextureLaws ("es", 2, 5);
  mean2 = text2.MeanImage (21, 21);

  histo = region.Histo2dim (mean1, mean2);
  feats = histo.Threshold (1.0, 1000000.0);

  win.SetDraw ("fill");
  win.SetColor ("red");

  feats.Display (win);

  cout << "Charakteristics area in red" << endl;

  cd2reg = mean1.Class2dimSup (mean2, feats);

  win.SetColor ("blue");
  cd2reg.Display (win);

  cout << "Result of classification in blue " << endl;
  win.Click ();
  return (0);
}

Let A be the area of the input region. Then the runtime complexity is O(256^2 + A).

class_2dim_sup returns 2 (H_MSG_TRUE). If all parameters are correct, the behavior with respect to the input images and output regions can be determined by setting the values of the flags 'no_object_result', 'empty_region_result', and 'store_empty_region' with set_system. If necessary, an exception is raised.

class_2dim_sup is reentrant and automatically parallelized (on tuple level).

histo_2dim, threshold, draw_region, dilation1, opening, shape_trans

connection, select_shape, select_gray

class_ndim_norm, class_ndim_box, threshold

histo_2dim

Foundation