RTK
2.6.0
Reconstruction Toolkit
|
#include <rtkGetNewtonUpdateImageFilter.h>
Public Types | |
using | dataType = typename TGradient::PixelType::ValueType |
using | Pointer = itk::SmartPointer< Self > |
using | Self = GetNewtonUpdateImageFilter |
using | Superclass = itk::ImageToImageFilter< TGradient, TGradient > |
Public Member Functions | |
virtual ::itk::LightObject::Pointer | CreateAnother () const |
virtual const char * | GetNameOfClass () const |
void | SetInputGradient (const TGradient *gradient) |
void | SetInputHessian (const THessian *hessian) |
Static Public Member Functions | |
static Pointer | New () |
Static Public Attributes | |
static constexpr unsigned int | nChannels = TGradient::PixelType::Dimension |
Protected Member Functions | |
void | DynamicThreadedGenerateData (const typename TGradient::RegionType &outputRegionForThread) override |
void | GenerateInputRequestedRegion () override |
GetNewtonUpdateImageFilter () | |
~GetNewtonUpdateImageFilter () override=default | |
TGradient::ConstPointer | GetInputGradient () |
THessian::ConstPointer | GetInputHessian () |
Computes update from gradient and Hessian in Newton's method.
This filter takes in inputs the gradient G (input 1) and the Hessian H (input 2) of a cost function, and computes the update U (output), by U = H^{-1} * G. In Newton's method, the quantity to add to the current iterate in order to get the next iterate is actually -U, so the minus operation has to be handled downstream. It is assumed that the cost function is separable, so that each pixel can be processed independently and has its own small G, H and U
Definition at line 45 of file rtkGetNewtonUpdateImageFilter.h.
using rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::dataType = typename TGradient::PixelType::ValueType |
Convenient type alias
Definition at line 70 of file rtkGetNewtonUpdateImageFilter.h.
using rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::Pointer = itk::SmartPointer<Self> |
Definition at line 53 of file rtkGetNewtonUpdateImageFilter.h.
using rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::Self = GetNewtonUpdateImageFilter |
Standard class type alias.
Definition at line 51 of file rtkGetNewtonUpdateImageFilter.h.
using rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::Superclass = itk::ImageToImageFilter<TGradient, TGradient> |
Definition at line 52 of file rtkGetNewtonUpdateImageFilter.h.
|
protected |
|
overrideprotecteddefault |
|
virtual |
Reimplemented from itk::Object.
|
overrideprotected |
Does the real work.
|
overrideprotectedvirtual |
Reimplemented from itk::ProcessObject.
|
protected |
Getters for the inputs
|
protected |
Getters for the inputs
|
virtual |
Run-time type information (and related methods).
Reimplemented from itk::ImageSource< TOutputImage >.
|
static |
Method for creation through the object factory.
void rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::SetInputGradient | ( | const TGradient * | gradient | ) |
Set methods for all inputs, since they have different types
void rtk::GetNewtonUpdateImageFilter< TGradient, THessian >::SetInputHessian | ( | const THessian * | hessian | ) |
Set methods for all inputs, since they have different types
|
static |
Convenient parameters extracted from template types
Definition at line 67 of file rtkGetNewtonUpdateImageFilter.h.