Use ParallelizeImageRegion

Synopsis

This example demonstrates how to use MultiThreaderBase::ParallelizeImageRegion to apply a non-trivial operation for all pixels in an image, in parallel. To perform a multi-threaded operation like this prior ITK 5.0 required the creation of a class.

Results

Segmentation and custom colorization of an image using MultiThreaderBase::ParallelizeImageRegion.

Code

C++

#include "itkImageFileReader.h"
#include "itkImageFileWriter.h"
#include "itkWatershedImageFilter.h"
#include "itkImageRegionIterator.h"

constexpr unsigned int Dimension = 2;
using InputPixelType = unsigned char;
using InputImageType = itk::Image<InputPixelType, Dimension>;
using OutputPixelType = itk::RGBAPixel<unsigned char>;
using OutputImageType = itk::Image<OutputPixelType, Dimension>;
using LabeledImageType = itk::Image<itk::IdentifierType, Dimension>;

OutputImageType::Pointer
segmentationAndCustomColorization(InputImageType::Pointer inImage)
{
  using WatershedFilterType = itk::WatershedImageFilter<InputImageType>;
  WatershedFilterType::Pointer watershed = WatershedFilterType::New();
  watershed->SetThreshold(0.05);
  watershed->SetLevel(0.3);
  watershed->SetInput(inImage);
  watershed->Update();

  LabeledImageType::Pointer image = watershed->GetOutput();
  image->DisconnectPipeline();

  OutputImageType::Pointer outImage = OutputImageType::New();
  outImage->CopyInformation(image);
  outImage->SetRegions(image->GetBufferedRegion());
  outImage->Allocate(true);

  itk::MultiThreaderBase::Pointer mt = itk::MultiThreaderBase::New();
  // ParallelizeImageRegion invokes the provided lambda function in parallel
  // each invocation will contain a piece of the overall region
  mt->ParallelizeImageRegion<Dimension>(
    image->GetBufferedRegion(),
    // the lambda will have access to outer variables 'image' and 'outImage'
    // it will have parameter 'region', which needs to be processed
    [image, outImage](const LabeledImageType::RegionType & region) {
      itk::ImageRegionConstIterator<LabeledImageType> iIt(image, region);
      itk::ImageRegionIterator<OutputImageType>       oIt(outImage, region);
      for (; !iIt.IsAtEnd(); ++iIt, ++oIt)
      {
        LabeledImageType::IndexType ind = iIt.GetIndex();
        OutputPixelType             p;
        p.SetAlpha(iIt.Get());
        static_assert(Dimension <= 3, "Dimension has to be 2 or 3");
        for (unsigned d = 0; d < Dimension; d++)
        {
          p.SetElement(d, ind[d]);
        }
        oIt.Set(p);
      }
    },
    nullptr); // we don't have a filter whose progress needs to be updated

  return outImage;
}

int
main(int argc, char * argv[])
{
  if (argc != 3)
  {
    std::cerr << "Usage: " << std::endl;
    std::cerr << argv[0];
    std::cerr << " <InputFileName>";
    std::cerr << " <OutputFileName>" << std::endl;
    return EXIT_FAILURE;
  }

  using ReaderType = itk::ImageFileReader<InputImageType>;
  ReaderType::Pointer reader = ReaderType::New();
  reader->SetFileName(argv[1]);
  using WriterType = itk::ImageFileWriter<OutputImageType>;
  WriterType::Pointer writer = WriterType::New();
  writer->SetFileName(argv[2]);
  writer->UseCompressionOn();

  try
  {
    OutputImageType::Pointer outImage = segmentationAndCustomColorization(reader->GetOutput());
    writer->SetInput(outImage);
    writer->Update();
  }
  catch (itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  return EXIT_SUCCESS;
}

Classes demonstrated

class MultiThreaderBase : public itk::Object

A class for performing multithreaded execution.

Multithreaders are a class hierarchy that provides support for multithreaded execution by abstracting away platform-specific details. This class can be used to execute a single method on multiple threads or to parallelize an operation over a given image region or array.

Subclassed by itk::PoolMultiThreader, itk::TBBMultiThreader

See itk::MultiThreaderBase for additional documentation.