Singlephase Chan and Vese Dense Field Level Set Segmentation#
Note
Wish List Still needs additional work to finish proper creation of example.
Synopsis#
Single-phase Chan And Vese Dense Field Level Set Segmentation
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++#
// The use of the ScalarChanAndVeseDenseLevelSetImageFilter is
// illustrated in the following example. The implementation of this filter in
// ITK is based on the paper by Chan And Vese. This
// implementation extends the functionality of the
// level-set filters in ITK by using region-based variational techniques. These methods
// do not rely on the presence of edges in the images.
//
// ScalarChanAndVeseDenseLevelSetImageFilter expects two inputs. The first is
// an initial level set in the form of an \doxygen{Image}. The second input
// is a feature image. For this algorithm, the feature image is the original
// raw or preprocessed image. Several parameters are required by the algorithm
// for regularization and weights of different energy terms. The user is encouraged to
// change different parameter settings to optimize the code example on their images.
//
// Let's start by including the headers of the main filters involved in the
// preprocessing.
//
#include "itkScalarChanAndVeseDenseLevelSetImageFilter.h"
#include "itkScalarChanAndVeseLevelSetFunctionData.h"
#include "itkConstrainedRegionBasedLevelSetFunctionSharedData.h"
#include "itkFastMarchingImageFilter.h"
#include "itkImageFileReader.h"
#include "itkImageFileWriter.h"
#include "itkImage.h"
#include "itkAtanRegularizedHeavisideStepFunction.h"
int
main(int argc, char ** argv)
{
if (argc < 6)
{
std::cerr << "Missing arguments" << std::endl;
std::cerr << "Usage: " << std::endl;
std::cerr << argv[0] << " featureImage outputImage";
std::cerr << " startx starty seedValue" << std::endl;
return EXIT_FAILURE;
}
unsigned int nb_iteration = 500;
double rms = 0.;
double epsilon = 1.;
double curvature_weight = 0.;
double area_weight = 0.;
double reinitialization_weight = 0.;
double volume_weight = 0.;
double volume = 0.;
double l1 = 1.;
double l2 = 1.;
//
// We now define the image type using a particular pixel type and
// dimension. In this case the \code{float} type is used for the pixels
// due to the requirements of the smoothing filter.
//
constexpr unsigned int Dimension = 2;
using ScalarPixelType = float;
using InternalImageType = itk::Image<ScalarPixelType, Dimension>;
using DataHelperType = itk::ScalarChanAndVeseLevelSetFunctionData<InternalImageType, InternalImageType>;
using SharedDataHelperType =
itk::ConstrainedRegionBasedLevelSetFunctionSharedData<InternalImageType, InternalImageType, DataHelperType>;
using LevelSetFunctionType =
itk::ScalarChanAndVeseLevelSetFunction<InternalImageType, InternalImageType, SharedDataHelperType>;
// We declare now the type of the numerically discretized Step and Delta functions that
// will be used in the level-set computations for foreground and background regions
//
using DomainFunctionType = itk::AtanRegularizedHeavisideStepFunction<ScalarPixelType, ScalarPixelType>;
auto domainFunction = DomainFunctionType::New();
domainFunction->SetEpsilon(epsilon);
InternalImageType::Pointer featureImage = itk::ReadImage<InternalImageType>(argv[1]);
// We declare now the type of the FastMarchingImageFilter that
// will be used to generate the initial level set in the form of a distance
// map.
//
using FastMarchingFilterType = itk::FastMarchingImageFilter<InternalImageType, InternalImageType>;
auto fastMarching = FastMarchingFilterType::New();
// The FastMarchingImageFilter requires the user to provide a seed
// point from which the level set will be generated. The user can actually
// pass not only one seed point but a set of them. Note the the
// FastMarchingImageFilter is used here only as a helper in the
// determination of an initial level set. We could have used the
// \doxygen{DanielssonDistanceMapImageFilter} in the same way.
//
// The seeds are passed stored in a container. The type of this
// container is defined as \code{NodeContainer} among the
// FastMarchingImageFilter traits.
//
using NodeContainer = FastMarchingFilterType::NodeContainer;
using NodeType = FastMarchingFilterType::NodeType;
auto seeds = NodeContainer::New();
InternalImageType::IndexType seedPosition;
seedPosition[0] = std::stoi(argv[3]);
seedPosition[1] = std::stoi(argv[4]);
const double initialDistance = std::stod(argv[5]);
NodeType node;
const double seedValue = -initialDistance;
node.SetValue(seedValue);
node.SetIndex(seedPosition);
// The list of nodes is initialized and then every node is inserted using
// the \code{InsertElement()}.
//
seeds->Initialize();
seeds->InsertElement(0, node);
// The set of seed nodes is passed now to the
// FastMarchingImageFilter with the method
// \code{SetTrialPoints()}.
//
fastMarching->SetTrialPoints(seeds);
// Since the FastMarchingImageFilter is used here just as a
// Distance Map generator. It does not require a speed image as input.
// Instead the constant value $1.0$ is passed using the
// \code{SetSpeedConstant()} method.
//
fastMarching->SetSpeedConstant(1.0);
// The FastMarchingImageFilter requires the user to specify the
// size of the image to be produced as output. This is done using the
// \code{SetOutputSize()}. Note that the size is obtained here from the
// output image of the smoothing filter. The size of this image is valid
// only after the \code{Update()} methods of this filter has been called
// directly or indirectly.
//
fastMarching->SetOutputSize(featureImage->GetBufferedRegion().GetSize());
fastMarching->Update();
// We declare now the type of the ScalarChanAndVeseDenseLevelSetImageFilter that
// will be used to generate a segmentation.
//
using MultiLevelSetType = itk::ScalarChanAndVeseDenseLevelSetImageFilter<InternalImageType,
InternalImageType,
InternalImageType,
LevelSetFunctionType,
SharedDataHelperType>;
auto levelSetFilter = MultiLevelSetType::New();
// We set the function count to 1 since a single level-set is being evolved.
//
levelSetFilter->SetFunctionCount(1);
// Set the feature image and initial level-set image as output of the
// fast marching image filter.
//
levelSetFilter->SetFeatureImage(featureImage);
levelSetFilter->SetLevelSet(0, fastMarching->GetOutput());
// Once activiated the level set evolution will stop if the convergence
// criteria or if the maximum number of iterations is reached. The
// convergence criteria is defined in terms of the root mean squared (RMS)
// change in the level set function. The evolution is said to have
// converged if the RMS change is below a user specified threshold. In a
// real application is desirable to couple the evolution of the zero set
// to a visualization module allowing the user to follow the evolution of
// the zero set. With this feedback, the user may decide when to stop the
// algorithm before the zero set leaks through the regions of low gradient
// in the contour of the anatomical structure to be segmented.
//
levelSetFilter->SetNumberOfIterations(nb_iteration);
levelSetFilter->SetMaximumRMSError(rms);
// Often, in real applications, images have different pixel resolutions. In such
// cases, it is best to use the native spacings to compute derivatives etc rather
// than sampling the images.
//
levelSetFilter->SetUseImageSpacing(1);
// For large images, we may want to compute the level-set over the initial supplied
// level-set image. This saves a lot of memory.
//
levelSetFilter->SetInPlace(false);
// For the level set with phase 0, set different parameters and weights. This may
// to be set in a loop for the case of multiple level-sets evolving simultaneously.
//
levelSetFilter->GetDifferenceFunction(0)->SetDomainFunction(domainFunction);
levelSetFilter->GetDifferenceFunction(0)->SetCurvatureWeight(curvature_weight);
levelSetFilter->GetDifferenceFunction(0)->SetAreaWeight(area_weight);
levelSetFilter->GetDifferenceFunction(0)->SetReinitializationSmoothingWeight(reinitialization_weight);
levelSetFilter->GetDifferenceFunction(0)->SetVolumeMatchingWeight(volume_weight);
levelSetFilter->GetDifferenceFunction(0)->SetVolume(volume);
levelSetFilter->GetDifferenceFunction(0)->SetLambda1(l1);
levelSetFilter->GetDifferenceFunction(0)->SetLambda2(l2);
levelSetFilter->Update();
try
{
itk::WriteImage(levelSetFilter->GetOutput(), argv[1]);
}
catch (const itk::ExceptionObject & excep)
{
std::cerr << "Exception caught !" << std::endl;
std::cerr << excep << std::endl;
return -1;
}
return EXIT_SUCCESS;
}
Classes demonstrated#
-
template<typename TInputImage, typename TFeatureImage, typename TOutputImage, typename TFunction = ScalarChanAndVeseLevelSetFunction<TInputImage, TFeatureImage>, class TSharedData = typename TFunction::SharedDataType>
class ScalarChanAndVeseDenseLevelSetImageFilter : public itk::MultiphaseDenseFiniteDifferenceImageFilter<TInputImage, TFeatureImage, TOutputImage, TFunction> Dense implementation of the Chan and Vese multiphase level set image filter.
This code was adapted from the paper:
"An active contour model without edges" T. Chan and L. Vese. In Scale-Space Theories in Computer Vision, pages 141-151, 1999.
This code was taken from the Insight Journal paper:
"Cell Tracking using Coupled Active Surfaces for Nuclei and Membranes" http://www.insight-journal.org/browse/publication/642 https://hdl.handle.net/10380/3055
- Author
Mosaliganti K., Smith B., Gelas A., Gouaillard A., Megason S.
That is based on the papers:
"Level Set Segmentation: Active Contours without edge" http://www.insight-journal.org/browse/publication/322 https://hdl.handle.net/1926/1532 and "Level set segmentation using coupled active surfaces" http://www.insight-journal.org/browse/publication/323 https://hdl.handle.net/1926/1533
