: this class contains a lot of functional code, which actually does not belong here.

Class elastix::AdaptiveStochasticGradientDescent< TElastix >

: this class contains a lot of functional code, which actually does not belong here.

Class elastix::AdaptiveStochasticLBFGS< TElastix >

: this class contains a lot of functional code, which actually does not belong here.

Class elastix::AdaptiveStochasticVarianceReducedGradient< TElastix >

: this class contains a lot of functional code, which actually does not belong here.

Class elastix::AdvancedBSplineTransform< TElastix >

It is unsure what happens when one of the image dimensions has length 1.

Class elastix::BSplineStackTransform< TElastix >

It is unsure what happens when one of the image dimensions has length 1.

Class elastix::BSplineTransformWithDiffusion< TElastix >

: this Transform has not been tested for images with Direction cosines matrix other than the identity matrix.

Global elastix::ELASTIX::RegisterImages (ImagePointer fixedImage, ImagePointer movingImage, const ParameterMapType &parameterMap, const std::string &outputPath, bool performLogging, bool performCout, ImagePointer fixedMask=nullptr, ImagePointer movingMask=nullptr)

generate file elastix_errors.h containing error codedefines (e.g. #define ELASTIX_NO_ERROR 0)

Global elastix::ElastixTemplate< TFixedImage, TMovingImage >::InputImageType

How useful is this? It is not consequently supported, since the the input image is stored in the MovingImageContainer anyway.

Class elastix::EulerStackTransform< TElastix >

It is unsure what happens when one of the image dimensions has length 1.

The center of rotation point is not transformed with the initial transform yet.

Global elastix::FullSearch< TElastix >::itkGetModifiableObjectMacro (OptimizationSurface, NDImageType)

BeforeAll, checking parameters.

Global elastix::MetricBase< TElastix >::m_ShowExactMetricValue

the method GetExactDerivative could as well be added here.

Class elastix::MultiBSplineTransformWithNormal< TElastix >

It is unsure what happens when one of the image dimensions has length 1.

Class elastix::PreconditionedStochasticGradientDescent< TElastix >

: this class contains a lot of functional code, which actually does not belong here.

Class elastix::RandomSamplerSparseMask< TElastix >

Write something similar for the RandomCoordinate sampler.

Class elastix::RecursiveBSplineTransform< TElastix >

It is unsure what happens when one of the image dimensions has length 1.

Global elastix::RegistrationBase< TElastix >::ReadMaskParameters (UseMaskErosionArrayType &useMaskErosionArray, const unsigned int nrOfMasks, const std::string &whichMask, const unsigned int level) const

: move to RegistrationBase Input:

Class itk::AdvancedImageMomentsCalculator< TImage >

It's not yet clear how multi-echo images should be handled here.

Class itk::AdvancedVersorTransform< TScalarType >

Need to make sure that the translation parameters in the baseclass cannot be set to non-zero values.

Class itk::FullSearchOptimizer

This optimizer has similar functionality as the recently added itkExhaustiveOptimizer. See if we can replace it by that optimizer, or inherit from it.

Global itk::KNNGraphAlphaMutualInformationImageToImageMetric< TFixedImage, TMovingImage >::EvaluateMovingFeatureImageDerivatives (const MovingImagePointType &mappedPoint, SpatialDerivativeType &featureGradients) const

move this to base class.

Class itk::NDImageBase< TPixel >

not all! An internal writer and reader are included for convenience.

Global itk::NDImageBase< TPixel >::CopyInformation (const DataObject *data)=0

Transform IndexToPoint methods.

Global itk::NDImageBase< TPixel >::SetRegions (SizeType size)=0

an NDRegionType should first be declared, in the same way as NDImage use SetRegions(size) for now. then knowlegde of the RegionType is not necessary. alternative: forget about the regiontype and add the functions SetLargestPossibleRegion, SetRegions etc with arguments (index,size) or maybe: ImageIORegion

: extend to direction cosines; but not needed for now in elastix

Global itk::NDImageTemplate< TPixel, VDimension >::CopyInformation (const DataObject *data) override

Transform IndexToPoint methods.

Global itk::ParzenWindowHistogramImageToImageMetric< TFixedImage, TMovingImage >::UpdateJointPDFAndIncrementalPDFs (RealType fixedImageValue, RealType movingImageValue, RealType movingMaskValue, const DerivativeType &movingImageValuesRight, const DerivativeType &movingImageValuesLeft, const DerivativeType &movingMaskValuesRight, const DerivativeType &movingMaskValuesLeft, const NonZeroJacobianIndicesType &nzji) const

The IsInsideMovingMask return bools are converted to doubles (1 or 0) to simplify the computation. But this may not be necessary.

Class itk::Statistics::ListSampleCArray< TMeasurementVector, TInternalValue >

: the second template argument should be removed, since the GetMeasurementVector method is incorrect when TMeasurementVector::ValueType != TInternalValue.

Global itk::WeightedCombinationTransform< TScalarType, NInputDimensions, NOutputDimensions >::TransformPointer

: shouldn't these be ConstPointers?

Class TranslationStackTransform

It is unsure what happens when one of the image dimensions has length 1.