Basic Region Growing#

Warning

Fix Problem Contains problems not fixed from original wiki.

Synopsis#

Basic region growing.

Results#

Note

Help Wanted Implementation of Results for sphinx examples containing this message. Reconfiguration of CMakeList.txt may be necessary. Write An Example <https://itk.org/ITKExamples/Documentation/Contribute/WriteANewExample.html>

Code#

C++#

#include "itkImage.h"
#include "itkRegionGrowImageFilter.h"
#include "itkCastImageFilter.h"

#include <itkImageToVTKImageFilter.h>

#include "vtkImageViewer.h"
#include "vtkRenderWindowInteractor.h"
#include "vtkSmartPointer.h"
#include "vtkImageActor.h"
#include "vtkInteractorStyleImage.h"
#include "vtkRenderer.h"

using PixelType = int;
constexpr size_t Dimension = 3;
using ImageType = itk::Image<PixelType, Dimension>;

static void
CreateImage(ImageType::Pointer image);

int
main(int argc, char * argv[])
{
  auto image = ImageType::New();
  CreateImage(image);

  using RegionGrowImageFilterType = itk::RegionGrowImageFilter<ImageType, ImageType>;
  auto  regionGrow = RegionGrowImageFilterType::New();
  float lower = 95.0;
  float upper = 105.0;
  regionGrow->SetLower(lower);
  regionGrow->SetUpper(upper);

  regionGrow->SetReplaceValue(255);

  // Seed 1: (25, 35)
  ImageType::IndexType seed1;
  seed1[0] = 25;
  seed1[1] = 35;
  regionGrow->SetSeed(seed1);
  regionGrow->SetInput(image);

  // Seed 2: (110, 120)
  ImageType::IndexType seed2;
  seed2[0] = 110;
  seed2[1] = 120;
  regionGrow->SetSeed(seed2);
  regionGrow->SetReplaceValue(150);

  regionGrow->SetInput(image);

  // Visualize
  using ConnectorType = itk::ImageToVTKImageFilter<ImageType>;
  auto connector2 = ConnectorType::New();
  connector2->SetInput(image2);

  vtkSmartPointer<vtkImageActor> actor2 = vtkSmartPointer<vtkImageActor>::New();
  actor2->SetInput(connector2->GetOutput());

  // Visualize joined image
  auto addConnector = ConnectorType::New();
  addConnector->SetInput(addFilter->GetOutput());

  vtkSmartPointer<vtkImageActor> addActor = vtkSmartPointer<vtkImageActor>::New();
  addActor->SetInput(addConnector->GetOutput());

  // There will be one render window
  vtkSmartPointer<vtkRenderWindow> renderWindow = vtkSmartPointer<vtkRenderWindow>::New();
  renderWindow->SetSize(900, 300);

  vtkSmartPointer<vtkRenderWindowInteractor> interactor = vtkSmartPointer<vtkRenderWindowInteractor>::New();
  interactor->SetRenderWindow(renderWindow);

  // Define viewport ranges
  // (xmin, ymin, xmax, ymax)
  double leftViewport[4] = { 0.0, 0.0, 0.33, 1.0 };
  double centerViewport[4] = { 0.33, 0.0, 0.66, 1.0 };
  double rightViewport[4] = { 0.66, 0.0, 1.0, 1.0 };

  // Setup both renderers
  vtkSmartPointer<vtkRenderer> leftRenderer = vtkSmartPointer<vtkRenderer>::New();
  renderWindow->AddRenderer(leftRenderer);
  leftRenderer->SetViewport(leftViewport);
  leftRenderer->SetBackground(.6, .5, .4);

  vtkSmartPointer<vtkRenderer> centerRenderer = vtkSmartPointer<vtkRenderer>::New();
  renderWindow->AddRenderer(centerRenderer);
  centerRenderer->SetViewport(centerViewport);
  centerRenderer->SetBackground(.4, .5, .6);

  vtkSmartPointer<vtkRenderer> rightRenderer = vtkSmartPointer<vtkRenderer>::New();
  renderWindow->AddRenderer(rightRenderer);
  rightRenderer->SetViewport(rightViewport);
  rightRenderer->SetBackground(.4, .5, .6);

  // Add the sphere to the left and the cube to the right
  leftRenderer->AddActor(actor1);
  centerRenderer->AddActor(actor2);
  rightRenderer->AddActor(addActor);

  leftRenderer->ResetCamera();
  centerRenderer->ResetCamera();
  rightRenderer->ResetCamera();

  renderWindow->Render();

  vtkSmartPointer<vtkInteractorStyleImage> style = vtkSmartPointer<vtkInteractorStyleImage>::New();
  interactor->SetInteractorStyle(style);

  interactor->Start();


  return EXIT_SUCCESS;
}

void
CreateImage(ImageType::Pointer image)
{
  // Create an image with 2 connected components
  ImageType::RegionType region;
  ImageType::IndexType  start;
  start[0] = 0;
  start[1] = 0;

  ImageType::SizeType size;
  size[0] = 200;
  size[1] = 300;

  region.SetSize(size);
  region.SetIndex(start);

  image->SetRegions(region);
  image->Allocate();

  // Make a square
  for (unsigned int r = 20; r < 80; ++r)
  {
    for (unsigned int c = 30; c < 100; ++c)
    {
      ImageType::IndexType pixelIndex;
      pixelIndex[0] = r;
      pixelIndex[1] = c;

      image->SetPixel(pixelIndex, 100.0);
    }
  }

  // Make another square
  for (unsigned int r = 100; r < 130; ++r)
  {
    for (unsigned int c = 115; c < 160; ++c)
    {
      ImageType::IndexType pixelIndex;
      pixelIndex[0] = r;
      pixelIndex[1] = c;

      image->SetPixel(pixelIndex, 100.0);
    }
  }
}

Classes demonstrated#

template<typename TInputImage, typename TOutputImage>
class RegionGrowImageFilter : public itk::ImageToImageFilter<TInputImage, TOutputImage>

Base class for RegionGrowImageFilter object.

itkRegionGrowImageFilter is the base class for the RegionGrowImageFilter objects. It provides the basic function definitions that are inherent to a RegionGrowImageFilter objects. It is templated over the type of input and output image.

This object defines the interface for those algorithm that perform feature/object segmentation by merging regions (parts of the image) that are similar in nature based on some metric. As a result parts of the image which belong to the same object gets merged and the region grows.

As an example regarding using this class to implementation of advanced region growing algorithm, itkRegionGrowImageFilterKLM class has been derived from this class. The virtual function ApplyRegionGrowImageFilter() provides the interface to the outside world to extend/enhance the scope of the current algorithm or write other region growing algorithms. The function MergeRegions is interface for the operation that merges two regions.

The local variable GridSize is used to define the initial small regions that the image is fragmented (atomic regions) into. For an 12 x 12 input image, GridSize set equal to [3, 3] will result in 16 initial regions. The default values are set equal to 2. The user can sets the number of desired regions via the m_MaxNumRegions parameter and the algorithm tries to perform region merging until there are only m_MaxNumRegions. If m_MaxNumRegions is more than the number of initial blocks, no region merging occurs.

These blocks are important as the labels associated with these blocks keep changing during the region growing process and at the end, while generating the results, each of these atomic blocks are revisited and the blocks with same labels are considered to belong to the same region.

This object supports data handling of multiband images. The object accepts images in vector format, where each pixel is a vector and each element of the vector corresponds to an entry from 1 particular band of a multiband dataset. The input to this object is assumed to be a multiband vector image, and the output is defined by specific algorithm implementation. The second template parameter is used to generate the the output image and can be modified according the algorithm specific output type.

We expect the user to provide the input to the routine in vector format. A single band image is treated as a vector image with a single element for every vector.

ITK Sphinx Examples:

Subclassed by itk::KLMRegionGrowImageFilter< TInputImage, TOutputImage >

See itk::RegionGrowImageFilter for additional documentation.