This class implements an extension of the weighted point based registration algorithm from A. Danilchenko, R. Balachandran and J. M. Fitzpatrick. More...

#include <mitkWeightedPointTransform.h>

Inheritance diagram for mitk::WeightedPointTransform:

[legend]

Collaboration diagram for mitk::WeightedPointTransform:

[legend]

Public Member Functions
	mitkClassMacroItkParent (WeightedPointTransform, itk::Object)

Pointer	Clone () const

void	ComputeTransformation ()
	Method which registers both point sets. More...

virtual void	SetThreshold (double _arg)
	Sets the threshold of the registration. Default value is 0.0001. More...

virtual void	SetMaxIterations (double _arg)
	Sets the maximum number of iterations of the registration. Default value is 1000. More...

virtual int	GetIterations ()

virtual double	GetFRE ()

virtual void	SetFRENormalizationFactor (double _arg)
	Sets the FRE normalization factor. Default value is 1.0. More...

virtual double	GetFRENormalizationFactor ()

void	SetMovingPointSet (vtkSmartPointer< vtkPoints > p)

void	SetCovarianceMatricesMoving (const CovarianceMatrixList &matrices)

void	SetFixedPointSet (vtkSmartPointer< vtkPoints > p)

void	SetCovarianceMatricesFixed (const CovarianceMatrixList &matrices)

const Translation &	GetTransformT () const

const Rotation &	GetTransformR () const

Static Public Member Functions
static Pointer	New ()

Protected Member Functions
	WeightedPointTransform ()

	~WeightedPointTransform () override

void	C_maker (vtkPoints *X, const WeightMatrixList &W, itk::VariableSizeMatrix< double > &returnValue)

void	E_maker (vtkPoints X, vtkPoints Y, const WeightMatrixList &W, vnl_vector< double > &returnValue)

double	CalculateConfigChange (vtkPoints X, vtkPoints X_new)

void	WeightedPointRegister (vtkPoints X, vtkPoints Y, const CovarianceMatrixList &Sigma_X, const CovarianceMatrixList &Sigma_Y, double Threshold, int MaxIterations, Rotation &TransformationR, Translation &TransformationT, double &FRE, int &n)
	This method performs a variant of the weighted point register algorithm presented by A. Danilchenko, R. Balachandran and J. M. Fitzpatrick in January 2010. (Modified in January 2011) converted to C++ by Alfred Franz in March/April 2010. More...

Protected Attributes
double	m_Threshold

int	m_MaxIterations

int	m_Iterations

double	m_FRE

double	m_FRENormalizationFactor

vtkSmartPointer< vtkLandmarkTransform >	m_LandmarkTransform

vtkSmartPointer< vtkPoints >	m_FixedPointSet

vtkSmartPointer< vtkPoints >	m_MovingPointSet

CovarianceMatrixList	m_CovarianceMatricesMoving

CovarianceMatrixList	m_CovarianceMatricesFixed

Translation	m_Translation

Rotation	m_Rotation

Detailed Description

This class implements an extension of the weighted point based registration algorithm from A. Danilchenko, R. Balachandran and J. M. Fitzpatrick.

The class implements an extension of the weighted point based registration from A. Danilchenko et al. presented by L. Maier-Hein et al. in "Convergent Iterative Closest-Point Algorithm to Accomodate Anisotropic and Inhomogenous Localization Error.", IEEE T Pattern Anal 34 (8), 1520-1532, 2012. The extension computes, in order to ensure the convergence of the algorithm, an isotropic estimation by an unweighted point based registration algorithm as an initial estimate. The implemantion was originally ported to C/C++ by A. Franz.

Note: Some methods are accelerated when OpenMP is enabled.

Definition at line 52 of file mitkWeightedPointTransform.h.

Constructor & Destructor Documentation

◆ WeightedPointTransform()

mitk::WeightedPointTransform::WeightedPointTransform ( )

protected

Definition at line 47 of file mitkWeightedPointTransform.cpp.

◆ ~WeightedPointTransform()

mitk::WeightedPointTransform::~WeightedPointTransform ( )

overrideprotected

Definition at line 57 of file mitkWeightedPointTransform.cpp.

References m_FixedPointSet, m_LandmarkTransform, and m_MovingPointSet.

Member Function Documentation

◆ C_maker()

void mitk::WeightedPointTransform::C_maker	(	vtkPoints *	X,
		const WeightMatrixList &	W,
		itk::VariableSizeMatrix< double > &	returnValue
	)

protected

original matlab-function:

Constructs the C matrix of the linear version of the registration problem, Cq = e, where q = [delta_angle(1:3),delta_translation(1:3)] and e is produced by e_maker(X,Y,W)

Authors: JM Fitzpatrick and R Balachandran Creation: February 2009

converted to C++ by Alfred Franz in March/April 2010

Definition at line 164 of file mitkWeightedPointTransform.cpp.

Referenced by WeightedPointRegister().

◆ CalculateConfigChange()

double mitk::WeightedPointTransform::CalculateConfigChange	(	vtkPoints *	X,
		vtkPoints *	X_new
	)

protected

This method computes the change in a root mean squared sense between the previous and the actual iteration. The computed value is used as a termination constraint of the algorithm and compared against the threshold.

Parameters

X	The moving point set in the previous iteration step.
X_new	The moving point set in the actual step.

Returns: The computed change between the two point sets.

Definition at line 414 of file mitkWeightedPointTransform.cpp.

References ls.

Referenced by WeightedPointRegister().

◆ Clone()

Pointer mitk::WeightedPointTransform::Clone ( ) const

◆ ComputeTransformation()

void mitk::WeightedPointTransform::ComputeTransformation ( )

Method which registers both point sets.

Definition at line 64 of file mitkWeightedPointTransform.cpp.

References m_CovarianceMatricesFixed, m_CovarianceMatricesMoving, m_FixedPointSet, m_FRE, m_Iterations, m_MaxIterations, m_MovingPointSet, m_Rotation, m_Threshold, m_Translation, and WeightedPointRegister().

◆ E_maker()

void mitk::WeightedPointTransform::E_maker	(	vtkPoints *	X,
		vtkPoints *	Y,
		const WeightMatrixList &	W,
		vnl_vector< double > &	returnValue
	)

protected

original matlab-function:

Constructs the e vector of the linear version of the registration problem, Cq = e, where q = [delta_angle(1:3),delta_translation(1:3)] and C is produced by C_maker(X,W)

Authors: JM Fitzpatrick and R Balachandran Creation: February 2009

converted to C++ by Alfred Franz in March/April 2010

Definition at line 188 of file mitkWeightedPointTransform.cpp.

Referenced by WeightedPointRegister().

◆ GetFRE()

virtual double mitk::WeightedPointTransform::GetFRE ( )

virtual

Returns: Returns the FRE of the last run of the registration algorithm. Returns -1 if there was no run of the registration yet.

◆ GetFRENormalizationFactor()

virtual double mitk::WeightedPointTransform::GetFRENormalizationFactor ( )

virtual

Returns: Returns the current FRE normalization factor.

◆ GetIterations()

virtual int mitk::WeightedPointTransform::GetIterations ( )

virtual

Returns: Returns the number of iterations of the last run of the registration algorithm. Returns -1 if there was no run of the registration yet.

◆ GetTransformR()

const Rotation& mitk::WeightedPointTransform::GetTransformR ( ) const

inline

The rotation matrix computed by the algorithm.

Definition at line 130 of file mitkWeightedPointTransform.h.

◆ GetTransformT()

const Translation& mitk::WeightedPointTransform::GetTransformT ( ) const

inline

The translation vector computed by the algorithm.

Returns: 3x1 translation vector.

Definition at line 126 of file mitkWeightedPointTransform.h.

◆ mitkClassMacroItkParent()

mitk::WeightedPointTransform::mitkClassMacroItkParent	(	WeightedPointTransform	,
		itk::Object
	)

◆ New()

static Pointer mitk::WeightedPointTransform::New ( )

static

◆ SetCovarianceMatricesFixed()

void mitk::WeightedPointTransform::SetCovarianceMatricesFixed ( const CovarianceMatrixList & matrices )

Set the list of 3x3 covariance matrices belonging to the fixed point set.

Parameters

matrices List of covariance matrices.

Definition at line 409 of file mitkWeightedPointTransform.cpp.

References m_CovarianceMatricesFixed.

◆ SetCovarianceMatricesMoving()

void mitk::WeightedPointTransform::SetCovarianceMatricesMoving ( const CovarianceMatrixList & matrices )

Set the list of 3x3 covariance matrices belonging to the moving point set.

Parameters

matrices List of covariance matrices.

Definition at line 399 of file mitkWeightedPointTransform.cpp.

References m_CovarianceMatricesMoving.

◆ SetFixedPointSet()

void mitk::WeightedPointTransform::SetFixedPointSet ( vtkSmartPointer< vtkPoints > p )

Sets the fixed point set used for the registration.

Parameters

p	The input point set.

Definition at line 404 of file mitkWeightedPointTransform.cpp.

References m_FixedPointSet.

◆ SetFRENormalizationFactor()

virtual void mitk::WeightedPointTransform::SetFRENormalizationFactor ( double _arg )

virtual

Sets the FRE normalization factor. Default value is 1.0.

◆ SetMaxIterations()

virtual void mitk::WeightedPointTransform::SetMaxIterations ( double _arg )

virtual

Sets the maximum number of iterations of the registration. Default value is 1000.

◆ SetMovingPointSet()

void mitk::WeightedPointTransform::SetMovingPointSet ( vtkSmartPointer< vtkPoints > p )

Sets the moving point set used for the registration.

Parameters

p	The input point set.

Definition at line 394 of file mitkWeightedPointTransform.cpp.

References m_MovingPointSet.

◆ SetThreshold()

virtual void mitk::WeightedPointTransform::SetThreshold ( double _arg )

virtual

Sets the threshold of the registration. Default value is 0.0001.

◆ WeightedPointRegister()

void mitk::WeightedPointTransform::WeightedPointRegister	(	vtkPoints *	X,
		vtkPoints *	Y,
		const CovarianceMatrixList &	Sigma_X,
		const CovarianceMatrixList &	Sigma_Y,
		double	Threshold,
		int	MaxIterations,
		Rotation &	TransformationR,
		Translation &	TransformationT,
		double &	FRE,
		int &	n
	)

protected

This method performs a variant of the weighted point register algorithm presented by A. Danilchenko, R. Balachandran and J. M. Fitzpatrick in January 2010. (Modified in January 2011) converted to C++ by Alfred Franz in March/April 2010.

Parameters

X	(input) the moving point set
Y	(input) the fixed (static) point set
Sigma_X	(input) a 3-by-3-by-N array, each page containing the weighting matrix for the Nth pair of points in X
Sigma_Y	(input) a 3-by-3-by-N array, each page containing the weighting matrix for the Nth pair of points in Y
Threshold	(input) the relative size of the change to the moving set above which the iteration continues
MaxIterations	(input) the maximum number of iterations allowed
Threshold	(input) the threshold used to terminate the algorithm
TransformationR	(output) this variable will hold the computed rotation matrix
TransformationT	(output) this variable will hold the computed translation vector
FRE	(output) this variable will hold the computed rotation FRE of the transformation
n	(output) this variable will hold the number of iterations used by the algorithm