ObjectiveFunction Class Reference

#include <ObjectiveFunction.h>

Public Types
typedef float	PrecisionType
typedef std::vector< std::vector< PrecisionType > >	SurrogateSignalType

Public Member Functions
	ObjectiveFunction ()
	~ObjectiveFunction ()
PrecisionType	GetValue (const PrecisionType *parametersIn)
void	GetGradient (const PrecisionType *parametersIn, PrecisionType *gradientOut, bool normaliseGradient=false)
std::vector< nifti_image * >	GetGradientAsImage (const PrecisionType *parametersIn, bool normaliseGradient=false)
unsigned int	GetNumberOfParameters ()
PrecisionType	GetMaxStepSize ()
void	SetCorrespondenceModel (const std::shared_ptr< CorrespondenceModel > &correspondenceModelIn)
void	SetSimilarityMeasure (const std::shared_ptr< ImageSimilarity > &imageSimilarityIn)
void	SetSurrogateSignals (const SurrogateSignalType &surrSignalsIn)
void	SetReferenceStateImage (nifti_image *refStateImgIn)
void	SetDynamicImages (const std::vector< nifti_image * > &dynamicImagesIn)
void	SetImageAcquisition (const std::shared_ptr< ImageAcquisition > &imageAcquisitionIn)

Private Attributes
std::shared_ptr< CorrespondenceModel >	correspondenceModel
	Object that represents the correspondence model. Able to generate a transformation. More...
std::shared_ptr< ImageSimilarity >	imageSimilarity
	Object that measures the similarity between two images. More...
SurrogateSignalType	surrogateSignals
	All surrogate signals. More...
nifti_image *	referenceStateImage
	Nifti image structure with all. More...
std::vector< nifti_image * >	dynamicImages
	Vector holding all pointers to the dynamic images. More...
PrecisionType	similarityWeight
	Similarity weight. More...
std::shared_ptr< ImageAcquisition >	imageAcquisition
	The object simulating acquisition and calculating the adjoint of the image acquisition procedure. More...

Detailed Description

Class to compute the objective function. The weigths of the similarity and penatly terms are computed internally.

Member Typedef Documentation

PrecisionType

typedef float ObjectiveFunction::PrecisionType

SurrogateSignalType

typedef std::vector< std::vector<PrecisionType> > ObjectiveFunction::SurrogateSignalType

Constructor & Destructor Documentation

ObjectiveFunction()

ObjectiveFunction::ObjectiveFunction

(

)

Constructor.

~ObjectiveFunction()

ObjectiveFunction::~ObjectiveFunction

(

)

Destructor.

Member Function Documentation

GetGradient()

void ObjectiveFunction::GetGradient	(	const PrecisionType *	parametersIn,
		PrecisionType *	gradientOut,
		bool	normaliseGradient = false
	)

Get the gradient of the objective function for a specific set of parameters This function copies the gradienbt of the objective function to an externally allocated memory location.

Parameters

parametersIn	Pointer to the parameters that will be fed into the correspondence model
gradientOut	Pointer to where the gradient will be written. Calling function needs to handle allocation of gradient.
normaliseGradient	Set to true if the gradient should be normalised with the maximum transformation length

GetGradientAsImage()

std::vector< nifti_image * > ObjectiveFunction::GetGradientAsImage	(	const PrecisionType *	parametersIn,
		bool	normaliseGradient = false
	)

Calculate the gradient for the input parameters and return an image with the gradient

Parameters

parametersIn	Pointer to the parameters that will be fed into the correspondence model
normaliseGradient	Set to true of the gradient should be normalised. Also see GetGradient().

GetMaxStepSize()

ObjectiveFunction::PrecisionType ObjectiveFunction::GetMaxStepSize

(

)

Get the maximum step size for the line-optimisation. Calcualtes the maximum step size \( l_\mathrm{max} \) from the voxel size of the reference-state image ( \( \{d_x, d_y, d_z\} \)) and the surrogate signal values \( s_i \).

\[ l_\mathrm{max} = \frac{\max(\{d_x, d_y, d_z\})}{\max( \| s_i \| ) } \]

Obsiously both, surrogate signal and reference-state image have to be set before calling this function.

GetNumberOfParameters()

unsigned int ObjectiveFunction::GetNumberOfParameters

(

)

Get the number of parameters (degrees of freedom) of the objectve function.

GetValue()

ObjectiveFunction::PrecisionType ObjectiveFunction::GetValue

(

const PrecisionType *

parametersIn

)

Get the value of the objective function for a specific set of parameters Calculates \( \mathcal{C}_\text{total} = \sum_{t=1}^{N_i} \lambda \mathcal{S}( \mathbf{P}_t, A_t(\mathbf{T}(\mathbf{I}, \mathbf{M}_t))) + (1-\lambda)\mathcal{R}(\mathbf{M}_t) \)

Parameters

parametersIn

Pointer to the parameters that will be fed into the correspondence model, or point \(\mathbf{R}\) in high-dimensional space for which the objective function-value will be calculated.

SetCorrespondenceModel()

void ObjectiveFunction::SetCorrespondenceModel

(

const std::shared_ptr< CorrespondenceModel > &

correspondenceModelIn

)

Set the correspondence model. When the correspondence model is set, the transformation is defined and the similarity and penalty term weights are updated.

Parameters

correspondenceModelIn

Pointer to the correspondence model used

SetDynamicImages()

void ObjectiveFunction::SetDynamicImages

(

const std::vector< nifti_image * > &

dynamicImagesIn

)

Set the dynamic images. Only set the resolution that is to be computed for each level.

Parameters

dynamicImagesIn	Vector holding the pointer to the dynamic images. Must be of same size as surrogate signals ObjectiveFunction::SetSurrogateSignals.
dynamicDataTypeIn	Defines the type of the dynamic data

SetImageAcquisition()

void ObjectiveFunction::SetImageAcquisition

(

const std::shared_ptr< ImageAcquisition > &

imageAcquisitionIn

)

Set the image acquisition object. Has to implement functionality to simulate the acquisition process and to calculate the adjoint of that process.

Parameters

imageAcquisitionIn

Shared pointer to the image acquisition object.

SetReferenceStateImage()

void ObjectiveFunction::SetReferenceStateImage

(

nifti_image *

refStateImgIn

)

Set the current reference state image

Parameters

refStateImgIn

Pointer to the nifti_image structure holding the reference state image

SetSimilarityMeasure()

void ObjectiveFunction::SetSimilarityMeasure

(

const std::shared_ptr< ImageSimilarity > &

imageSimilarityIn

)

Set the object that measures the similarity between images.

Parameters

imageSimilarityIn

Image similarity object.

SetSurrogateSignals()

void ObjectiveFunction::SetSurrogateSignals

(

const SurrogateSignalType &

surrSignalsIn

)

Set the surrogate signal To measure the objective function, all the surrogate values need to be known (and of correct size).

Member Data Documentation

correspondenceModel

std::shared_ptr<CorrespondenceModel> ObjectiveFunction::correspondenceModel

private

Object that represents the correspondence model. Able to generate a transformation.

dynamicImages

std::vector<nifti_image*> ObjectiveFunction::dynamicImages

private

Vector holding all pointers to the dynamic images.

imageAcquisition

std::shared_ptr<ImageAcquisition> ObjectiveFunction::imageAcquisition

private

The object simulating acquisition and calculating the adjoint of the image acquisition procedure.

imageSimilarity

std::shared_ptr<ImageSimilarity> ObjectiveFunction::imageSimilarity

private

Object that measures the similarity between two images.

referenceStateImage

nifti_image* ObjectiveFunction::referenceStateImage

private

Nifti image structure with all.

similarityWeight

PrecisionType ObjectiveFunction::similarityWeight

private

Similarity weight.

\[ 1-\lambda \]

surrogateSignals

SurrogateSignalType ObjectiveFunction::surrogateSignals

private

All surrogate signals.

---

The documentation for this class was generated from the following files:

• motion-lib/ObjectiveFunction.h
• motion-lib/ObjectiveFunction.cpp