Computes normalized cross correlation using a small neighborhood for each voxel between two images, with speed optimizations for dense registration. More...

#include <itkANTSNeighborhoodCorrelationImageToImageMetricv4.h>

Inheritance diagram for itk::ANTSNeighborhoodCorrelationImageToImageMetricv4< TFixedImage, TMovingImage, TVirtualImage >:

Collaboration diagram for itk::ANTSNeighborhoodCorrelationImageToImageMetricv4< TFixedImage, TMovingImage, TVirtualImage >:

List of all members.

Classes
struct	ScanMemType
struct	ScanParametersType
Public Types
typedef SmartPointer< const Self >	ConstPointer
typedef Superclass::DerivativeType	DerivativeType
typedef Superclass::DerivativeValueType	DerivativeValueType
typedef Superclass::FixedImageGradientType	FixedImageGradientType
typedef FixedTransformType::JacobianType	FixedImageJacobianType
typedef Superclass::FixedImagePixelType	FixedImagePixelType
typedef Superclass::FixedImagePointType	FixedImagePointType
typedef Superclass::FixedImageType	FixedImageType
typedef Superclass::FixedOutputPointType	FixedOutputPointType
typedef Superclass::FixedTransformType::JacobianType	FixedTransformJacobianType
typedef Superclass::FixedTransformType	FixedTransformType
typedef Superclass::ImageDimensionType	ImageDimensionType
typedef VirtualImageType::RegionType	ImageRegionType
typedef VirtualImageType::IndexType	IndexType
typedef Superclass::InternalComputationValueType	InternalComputationValueType
typedef Superclass::JacobianType	JacobianType
typedef Superclass::MeasureType	MeasureType
typedef Superclass::MovingImageGradientType	MovingImageGradientType
typedef MovingTransformType::JacobianType	MovingImageJacobianType
typedef Superclass::MovingImagePixelType	MovingImagePixelType
typedef Superclass::MovingImagePointType	MovingImagePointType
typedef Superclass::MovingImageType	MovingImageType
typedef Superclass::MovingOutputPointType	MovingOutputPointType
typedef Superclass::MovingTransformType::JacobianType	MovingTransformJacobianType
typedef Superclass::MovingTransformType	MovingTransformType
typedef Superclass::NumberOfParametersType	NumberOfParametersType
typedef SmartPointer< Self >	Pointer
typedef VirtualImageType::SizeType	RadiusType
typedef ANTSNeighborhoodCorrelationImageToImageMetricv4	Self
typedef ImageToImageMetricv4 < TFixedImage, TMovingImage, TVirtualImage >	Superclass
typedef Superclass::VirtualImageGradientType	VirtualImageGradientType
typedef Superclass::VirtualImageType	VirtualImageType
typedef Superclass::VirtualPointType	VirtualPointType
Public Member Functions
virtual ::itk::LightObject::Pointer	CreateAnother (void) const
virtual const char *	GetNameOfClass () const
virtual RadiusType	GetRadius ()
virtual RadiusType	GetRadius () const
virtual void	SetRadius (RadiusType _arg)
Static Public Member Functions
static Pointer	New ()
Static Public Attributes
static const ImageDimensionType	FixedImageDimension = ::itk::GetImageDimension<FixedImageType>::ImageDimension
static const ImageDimensionType	MovingImageDimension = ::itk::GetImageDimension<MovingImageType>::ImageDimension
static const ImageDimensionType	VirtualImageDimension = ::itk::GetImageDimension<VirtualImageType>::ImageDimension
Protected Types
typedef ANTSNeighborhoodCorrelationImageToImageMetricv4DenseGetValueAndDerivativeThreader < Superclass, Self >	ANTSNeighborhoodCorrelationImageToImageMetricv4DenseGetValueAndDerivativeThreaderType
typedef InternalComputationValueType	QueueRealType
typedef ConstNeighborhoodIterator < VirtualImageType >	ScanIteratorType
typedef struct itk::ANTSNeighborhoodCorrelationImageToImageMetricv4::ScanMemType	ScanMemType
typedef struct itk::ANTSNeighborhoodCorrelationImageToImageMetricv4::ScanParametersType	ScanParametersType
typedef std::deque< QueueRealType >	SumQueueType
Protected Member Functions
	ANTSNeighborhoodCorrelationImageToImageMetricv4 ()
void	InitializeScanning (const ImageRegionType &scanRegion, ScanIteratorType &scanIt, ScanMemType &scanMem, ScanParametersType &scanParameters) const
virtual void	PrintSelf (std::ostream &os, Indent indent) const
virtual	~ANTSNeighborhoodCorrelationImageToImageMetricv4 ()
Private Member Functions
	ANTSNeighborhoodCorrelationImageToImageMetricv4 (const Self &)
void	operator= (const Self &)
Private Attributes
RadiusType	m_Radius
Friends
class	ANTSNeighborhoodCorrelationImageToImageMetricv4DenseGetValueAndDerivativeThreader< Superclass, Self >

Detailed Description

template<class TFixedImage, class TMovingImage, class TVirtualImage = TFixedImage>
class itk::ANTSNeighborhoodCorrelationImageToImageMetricv4< TFixedImage, TMovingImage, TVirtualImage >

Computes normalized cross correlation using a small neighborhood for each voxel between two images, with speed optimizations for dense registration.

Please cite this reference for more details:

Brian B. Avants, Nicholas J. Tustison, Gang Song, Philip A. Cook, Arno Klein, James C. Gee, A reproducible evaluation of ANTs similarity metric performance in brain image registration, NeuroImage, Volume 54, Issue 3, 1 February 2011, Pages 2033-2044, ISSN 1053-8119, DOI: 10.1016/j.neuroimage.2010.09.025.

Around each voxel, the neighborhood is defined as a N-Dimensional rectangle centered at the voxel. The size of the rectangle is 2*radius+1. The normalized correlation between neighborhoods of fixed image and moving image are averaged over the whole image as the final metric.

This class uses a specific fast implementation that is described in the above paper. There are two particular speed-ups:

1) It is assumed that the derivative is only affected by changes in the transform at the center of the window. This is obviously not true but speeds the evaluation up considerably and works well in practice. This assumption is the main differentiation of this approach from a more generic one.

2) The evaluation uses on-the-fly queues with multi-threading and a sliding neighborhood window. This is described in the above paper and specifically optimized for dense registration.

Example of usage:

typedef itk::ANTSNeighborhoodCorrelationImageToImageMetricv4 <ImageType, ImageType> MetricType; typedef MetricType::Pointer MetricTypePointer; MetricTypePointer metric = MetricType::New();

// set all parameters Size<Dimension> neighborhoodRadius; neighborhoodRadius.Fill(2); metric->SetRadius(neighborhood_radius); metric->SetFixedImage(fixedImage); metric->SetMovingImage(movingImage); metric->SetFixedTransform(transformFix); metric->SetMovingTransform(transformMov);

// initialization after parameters are set. metric->Initialize();

// getting derivative and metric value metric->GetValueAndDerivative(valueReturn, derivativeReturn);

This class is templated over the type of the two input objects. This is the base class for a hierarchy of similarity metrics that may, in derived classes, operate on meshes, images, etc. This class computes a value that measures the similarity between the two objects.

Definition at line 92 of file itkANTSNeighborhoodCorrelationImageToImageMetricv4.h.