Main Page Groups Namespace List Class Hierarchy Alphabetical List Compound List File List Namespace Members Compound Members File Members Concepts

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage > Class Template Reference
[Level Set-Based Segmentation Filters]

#include <itkFiniteDifferenceImageFilter.h>

Inheritance diagram for itk::FiniteDifferenceImageFilter:

[legend]Collaboration diagram for itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >:

[legend]List of all members.

Public Types
typedef FiniteDifferenceImageFilter	Self
typedef ImageToImageFilter< TInputImage, TOutputImage >	Superclass
typedef TInputImage	InputImageType
typedef TOutputImage	OutputImageType
typedef TOutputImage::PixelType	PixelType
typedef FiniteDifferenceFunction< TOutputImage >	FiniteDifferenceFunctionType
typedef FiniteDifferenceFunctionType::TimeStepType	TimeStepType
Public Methods
virtual const char *	GetClassName () const
	itkStaticConstMacro (ImageDimension, unsigned int, OutputImageType::ImageDimension)
virtual unsigned int	GetElapsedIterations () const
	itkGetConstReferenceObjectMacro (DifferenceFunction, FiniteDifferenceFunctionType)
virtual void	SetDifferenceFunction (FiniteDifferenceFunctionType *_arg)
Protected Methods
	FiniteDifferenceImageFilter ()
	~FiniteDifferenceImageFilter ()
void	PrintSelf (std::ostream &os, Indent indent) const
virtual void	AllocateUpdateBuffer ()=0
virtual void	ApplyUpdate (TimeStepType dt)=0
virtual TimeStepType	CalculateChange ()=0
virtual void	CopyInputToOutput ()
virtual void	GenerateData ()
virtual void	GenerateInputRequestedRegion ()
virtual bool	Halt ()=0
virtual void	Initialize ()
virtual void	InitializeIteration ()
virtual TimeStepType	ResolveTimeStep (const TimeStepType list, const bool valid, int size)
virtual void	SetElapsedIterations (unsigned int _arg)
virtual void	PostProcessOutput ()

Detailed Description

template<class TInputImage, class TOutputImage>
class itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >

The Finite Difference Solver Hierarchy: This is an overview of the finite difference solver (FDS) framework. The FDS framework is a set of classes for creating filters to solve partial differential equations on images using an iterative, finite difference update scheme.
The high-level algorithm implemented by the framework can be described by the following pseudocode.

  WHILE NOT convergence:
     FOR ALL pixels i
        time_step = calculate_change(i)
        update(i, time_step)

The following equation describes update at pixel on discrete image :

Component objects: The FDS hierarchy is comprised of two component object types, variations of which are designed to be plugged together to create filters for different applications. At the process level are the ``solver'' objects, which are subclasses of FiniteDifferenceImageFilter. Solver objects are filters that take image inputs and produce image outputs. Solver objects require a ``finite difference function'' object to perform the calculation at each image pixel during iteration. These specialized function objects are subclasses of FiniteDifferenceFunction. FiniteDifferenceFunctions take a neighborhood of pixels as input (in the form of an itk::NeighborhoodIterator) and produce a scalar valued result.
Filters for different applications are created by defining a function object to handle the numerical calculations and choosing (or creating) a solver object that reflects the requirements and constraints of the application. For example, anisotropic diffusion filters are created by plugging anisotropic diffusion functions into the DenseFiniteDifferenceImageFilter. The separation between function object and solver object allows us to create, for example, sparse-field, dense-field, and narrow-band implementations of a level-set surface evolution filter can all be constructed by plugging the same function object into three different, specialized solvers.

Creating new filters in this hierarchy: The procedure for creating a filter within the FDS hierarchy is to identify all the virtual methods that need to be defined for your particular application. In the simplest case, a filter needs only to instantiate a specific function object and define some halting criteria. For more complicated applications, you may need to define a specialized type of iteration scheme or updating procedure in a higher-level solver object.
Some simple examples are the specific subclasses of AnisotropicDiffusionImageFilter. The leaves of the anisotropic diffusion filter tree only define the function object they use for their particular flavor of diffusion. See CurvatureAnisotropicDiffusionImageFilter and GradientAnisotropicDiffusionImageFilter for details.

FiniteDifferenceImageFilter: This class defines the generic solver API at the top level of the FDS framework. FiniteDifferenceImageFilter is an abstract class that implements the generic, high-level algorithm (described above).

Inputs and Outputs: This filter is an Image to Image filter. Depending on the specific subclass implementation, finite difference image filters may process a variety of image types. The input to the filter is the initial value of and the output of the filter is the solution to the p.d.e.

How to use this class: GenerateData() relies on several virtual methods that must be defined by a subclass. Specifically: AllocateUpdateBuffer ApplyUpdate CalculateChange and Halt. To create a finite difference solver, implement a subclass to define these methods.
Note that there is no fixed container type for the buffer used to hold the update . The container might be another image, or simply a list of values. AllocateUpdateBuffer is responsible for creating the container. CalculateChange populates this buffer and ApplyUpdate adds the buffer values to the output image (solution). The boolean Halt() method returns a true value to stop iteration.

See also:: DenseFiniteDifferenceImageFilter

Definition at line 122 of file itkFiniteDifferenceImageFilter.h.

Member Typedef Documentation

template<class TInputImage, class TOutputImage>

typedef FiniteDifferenceFunction<TOutputImage> itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::FiniteDifferenceFunctionType

The value type of the time step. This is distinct from PixelType because PixelType may often be a vector value, while the TimeStep is a scalar value.
Reimplemented in itk::BinaryMinMaxCurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::DemonsRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, itk::MinMaxCurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::PDEDeformableRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, itk::DenseFiniteDifferenceImageFilter< TInputImage, TOutputImage >, and itk::DenseFiniteDifferenceImageFilter< TDeformationField, TDeformationField >.
Definition at line 148 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

typedef TOutputImage::PixelType itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::PixelType

The pixel type of the output image will be used in computations.
Reimplemented in itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::AnisotropicDiffusionImageFilter< TInputImage, TOutputImage >, itk::DenseFiniteDifferenceImageFilter< TInputImage, TOutputImage >, and itk::DenseFiniteDifferenceImageFilter< TDeformationField, TDeformationField >.
Definition at line 143 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

typedef FiniteDifferenceFunctionType::TimeStepType itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::TimeStepType

Reimplemented in itk::AntiAliasBinaryImageFilter< TInputImage, TOutputImage >, itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::SegmentationLevelSetImageFilter< TInputImage, TOutputImage >, itk::AnisotropicDiffusionImageFilter< TInputImage, TOutputImage >, itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >, itk::DenseFiniteDifferenceImageFilter< TInputImage, TOutputImage >, and itk::DenseFiniteDifferenceImageFilter< TDeformationField, TDeformationField >.
Definition at line 149 of file itkFiniteDifferenceImageFilter.h.
Referenced by itk::FiniteDifferenceImageFilter< TDeformationField, TDeformationField >::InitializeIteration(), and itk::FiniteDifferenceImageFilter< TDeformationField, TDeformationField >::~FiniteDifferenceImageFilter().

Constructor & Destructor Documentation

template<class TInputImage, class TOutputImage>

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::FiniteDifferenceImageFilter ( ) [inline, protected]

Definition at line 166 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::~FiniteDifferenceImageFilter ( ) [inline, protected]

Definition at line 167 of file itkFiniteDifferenceImageFilter.h.

Member Function Documentation

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::AllocateUpdateBuffer ( ) [protected, pure virtual]

This method allocates a temporary update container in the subclass.
Implemented in itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::ApplyUpdate ( TimeStepType dt ) [protected, pure virtual]

This method is defined by a subclass to apply changes to the output from an update buffer and a time step value "dt".
Parameters:

dt Time step value.

Implemented in itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual TimeStepType itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::CalculateChange ( ) [protected, pure virtual]

This method is defined by a subclass to populate an update buffer with changes for the pixels in the output. It returns a time step value to be used for the update.
Returns:
A time step to use in updating the output with the changes calculated from this method.

Implemented in itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::CopyInputToOutput ( ) [protected, virtual]

A simple method to copy the data from the input to the output. ( Supports "read-only" image adaptors in the case where the input image type converts to a different output image type. )
Reimplemented in itk::PDEDeformableRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, and itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::GenerateData ( ) [protected, virtual]

This is the default, high-level algorithm for calculating finite difference solutions. It calls virtual methods in its subclasses to implement the major steps of the algorithm.
Reimplemented from itk::ImageSource< TOutputImage >.
Reimplemented in itk::AntiAliasBinaryImageFilter< TInputImage, TOutputImage >, and itk::SegmentationLevelSetImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::GenerateInputRequestedRegion ( ) [protected, virtual]

FiniteDifferenceImageFilter needs a larger input requested region than the output requested region. As such, we need to provide an implementation for GenerateInputRequestedRegion() in order to inform the pipeline execution model.
The filter will ask for a padded region to perform its neighborhood calculations. If no such region is available, the boundaries will be handled as described in the FiniteDifferenceFunction defined by the subclass.
See also:
ProcessObject::GenerateInputRequestedRegion()

Reimplemented from itk::ImageToImageFilter< TInputImage, TOutputImage >.
Reimplemented in itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, and itk::PDEDeformableRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >.

template<class TInputImage, class TOutputImage>

virtual unsigned int itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::GetElapsedIterations ( ) const [virtual]

Get the number of elapsed iterations of the filter.
Referenced by itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >::Halt().

template<class TInputImage, class TOutputImage>

virtual bool itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::Halt ( ) [protected, pure virtual]

This method returns true when the current iterative solution of the equation has met the criteria to stop solving. Defined by a subclass.
Implemented in itk::AntiAliasBinaryImageFilter< TInputImage, TOutputImage >, itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::PDEDeformableRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, itk::SegmentationLevelSetImageFilter< TInputImage, TOutputImage >, and itk::AnisotropicDiffusionImageFilter< TInputImage, TOutputImage >.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::Initialize ( void ) [inline, protected, virtual]

This method is optionally defined by a subclass and is called before the loop of iterations of calculate_change & upate. It does the global initialization, i.e. in the SparseFieldLevelSetImageFilter, initialize the list of layers.
Reimplemented in itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.
Definition at line 217 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::InitializeIteration ( ) [inline, protected, virtual]

This method is optionally defined by a subclass and is called immediately prior to each iterative CalculateChange-ApplyUpdate cycle. It can be used to set global variables needed for the next iteration (ie. average gradient magnitude of the image in anisotropic diffusion functions), or otherwise prepare for the next iteration.
Reimplemented in itk::BinaryMinMaxCurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::CurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::DemonsRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, itk::MinMaxCurvatureFlowImageFilter< TInputImage, TOutputImage >, itk::PDEDeformableRegistrationFilter< TFixedImage, TMovingImage, TDeformationField >, itk::SegmentationLevelSetImageFilter< TInputImage, TOutputImage >, and itk::AnisotropicDiffusionImageFilter< TInputImage, TOutputImage >.
Definition at line 225 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::itkGetConstReferenceObjectMacro ( DifferenceFunction ,

FiniteDifferenceFunctionType

)

This method returns a pointer to a FiniteDifferenceFunction object that will be used by the filter to calculate updates at image pixels.
Returns:
A FiniteDifferenceObject pointer.

template<class TInputImage, class TOutputImage>

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::itkStaticConstMacro ( ImageDimension ,

unsigned int,

OutputImageType::ImageDimension

)

Dimensionality of input and output data is assumed to be the same.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::PostProcessOutput ( ) [inline, protected, virtual]

This method is called after the solution has been generated to allow subclasses to apply some further processing to the output.
Reimplemented in itk::SparseFieldLevelSetImageFilter< TInputImage, TOutputImage >.
Definition at line 249 of file itkFiniteDifferenceImageFilter.h.

template<class TInputImage, class TOutputImage>

virtual TimeStepType itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::ResolveTimeStep ( const TimeStepType * list,

const bool * valid,

int size

) [protected, virtual]

Virtual method for resolving a single time step from a set of time steps returned from processing threads.
Returns:
Time step (dt) for the iteration update based on a list of time steps generated from the threaded calculated change method (one for each region processed).
Parameters:

list The set of time changes compiled from all the threaded calls to ThreadedGenerateData.

valid The set of flags indicating which of "list" elements are valid

size The size of "list" and "valid"

The default is to return the minimum value in the list.

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::SetDifferenceFunction ( FiniteDifferenceFunctionType * _arg ) [virtual]

This method sets the pointer to a FiniteDifferenceFunction object that will be used by the filter to calculate updates at image pixels.
Returns:
A FiniteDifferenceObject pointer.

Referenced by itk::CurvatureAnisotropicDiffusionImageFilter< TInputImage, TOutputImage >::CurvatureAnisotropicDiffusionImageFilter(), itk::GradientAnisotropicDiffusionImageFilter< TInputImage, TOutputImage >::GradientAnisotropicDiffusionImageFilter(), itk::VectorCurvatureAnisotropicDiffusionImageFilter< TInputImage, TOutputImage >::VectorCurvatureAnisotropicDiffusionImageFilter(), and itk::VectorGradientAnisotropicDiffusionImageFilter< TInputImage, TOutputImage >::VectorGradientAnisotropicDiffusionImageFilter().

template<class TInputImage, class TOutputImage>

virtual void itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage >::SetElapsedIterations ( unsigned int _arg ) [protected, virtual]

Set the number of elapsed iterations of the filter.

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

itkFiniteDifferenceImageFilter.h

Generated at Wed Mar 12 01:16:06 2003 for ITK by

itk::FiniteDifferenceImageFilter< TInputImage, TOutputImage > Class Template Reference [Level Set-Based Segmentation Filters]

Public Types

Public Methods