[Insight-users] Still struggling with extracting information about objects resulting from a segmentation

[Neal R. Harvey]

I am still struggling with getting the information I need about the 
various regions (objects) that
I obtain from a segmentation.
I am using ITK's watershed code to segment an image.
Then, I want to calculate various attributes of each of the regions 
found in the segmentation.
For attribute calculation, I am trying to use the ShapeLabelMapFilter, 
etc. as described in
Lehmann's paper, entitled "Label Object Representation and Manipulation 
with ITK".
Richard Beare offered some hints last time and I have been trying to use 
that, together with the
information in the paper referred to above.

Below is the snippet of code that I wrote that starts at the point where 
I have obtained a
segmentation from the watershed algorithm. The code compiles with no 
problem.
What I would expect that, if this was doing what I was hoping it would 
do, I would see
the same number of objects being reported from the ShapeLabelMapFilter 
as the number
of regions that is getting reported from the watershed segmentation. 
However, I get the
number of objects reported from the ShapeLabelMapFilter as zero.

Can anyone suggest a fix or has any idea as to what I am doing wrong?

On a slightly different, but related (i.e. it isn't a problem at the 
moment, as the problem
described above is not allowing the calculation of the object 
attributes, as it doesn't
think there are any object), is if I include calls to GetFeretDiameter() 
or GetEquivalentRadius()
the code won't even compile!

  WatershedFilter->SetLevel( watershedLevel );
  WatershedFilter->SetThreshold( watershedThreshold );
  WatershedFilter->SetInput( TemporaryImage );
  WatershedFilter->Update();

  // The output of the watershed is an image of unsigned long int

  typedef unsigned long       WatershedPixelType;

  typedef itk::Image< WatershedPixelType, Dimension >     
WatershedImageType;

  WatershedImageType::Pointer WatershedImage = WatershedFilter->GetOutput();

  //  Below we instantiate the MinimumMaximumImageCalculator class

  typedef itk::MinimumMaximumImageFilter< WatershedImageType > 
MinimumMaximumWatershedImageFilterType;

  // A MinimumMaximumWatershedImageFilter object is constructed

  MinimumMaximumWatershedImageFilterType::Pointer 
MinimumMaximumWatershedImageFilter = 
MinimumMaximumWatershedImageFilterType::New();

  MinimumMaximumWatershedImageFilter->SetInput( WatershedImage );

  MinimumMaximumWatershedImageFilter->Update();

  WatershedPixelType outputMinVal = 
MinimumMaximumWatershedImageFilter->GetMinimum();
  WatershedPixelType outputMaxVal = 
MinimumMaximumWatershedImageFilter->GetMaximum();

  std::cerr << "Minimum Value from Segmented Image = " << 
(int)outputMinVal << std::endl;
  std::cerr << "Maximum Value from Segmented Image = " << 
(int)outputMaxVal << std::endl;

  // OK So, now we have a labeled image output from the watershed
  // We now want to calculate various characteristics of the objects in the
  // labeled image

  typedef unsigned long LabelType;
  typedef itk::ShapeLabelObject< LabelType, Dimension > LabelObjectType;
  typedef itk::LabelMap< LabelObjectType > LabelMapType;

  // First we need to convert the output from the watershed filter to a
  // collection of label objects

  typedef itk::LabelImageToShapeLabelMapFilter< WatershedImageType, 
LabelMapType > ConverterType;
  ConverterType::Pointer converter = ConverterType::New();
  converter->SetInput( WatershedImage );

  // Then, can we evaluate the attributes with the dedicated filtr: 
ShapeLabelMapFilter?
  typedef itk::ShapeLabelMapFilter< LabelMapType > ShapeFilterType;
  ShapeFilterType::Pointer shape = ShapeFilterType::New();
  shape->SetInput( converter->GetOutput() );

  shape->Update();

  LabelMapType::Pointer labelMap = converter->GetOutput();

  unsigned int NumberOfLabelObjects = labelMap->GetNumberOfLabelObjects();

  std::cout << "labelMap->GetNumberOfLabelObjects() = " << 
NumberOfLabelObjects << std::endl;

  if (NumberOfLabelObjects < 1) {

    std::cerr << "Number of Label Objects is incorrect" << std::endl;

    return EXIT_FAILURE;

  } else {

    for (unsigned int label = 1; label <= NumberOfLabelObjects; label++) {
      const LabelObjectType * labelObject = labelMap->GetLabelObject( 
label );

      //"; FeretDiameter = " << labelObject->GetFeretDiameter() << "; 
EquivalentRadius = " << labelObject->GetEquivalentRadius()  <<

      std::cout << "Label #: " << label << "; Size = " << 
labelObject->GetSize() << "; Perimeter = " << 
labelObject->GetPerimeter() << std::endl;
     
      return EXIT_SUCCESS;

    }

  }