Skip to content
Snippets Groups Projects
Commit 6eb7f0ca authored by Julien Michel's avatar Julien Michel
Browse files

ENH: Handle lambdas returning VariableLengthVector

parent 8f1f7f3b
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Modules/Filtering/ImageManipulation/include/otbFunctorImageFilter.h
+ 41
− 23
View file @ 6eb7f0ca
......@@ -30,6 +30,7 @@
#include "itkImageRegionConstIterator.h"
#include "itkProcessObject.h"
#include <type_traits>
#include <utility>
#include "otbFunctionTraits.h"
#include "itkDefaultConvertPixelTraits.h"
......@@ -174,7 +175,7 @@ template <class F, class T, size_t N> struct NumberOfOutputComponents<F,otb::Ima
static void Set(const F&, otb::Image<T> *, std::array<size_t,N>){ std::cout<<"Default set"<<std::endl;}
};
// Case 1: O is a VectorImage AND F has a fixed OuptutSize static constexrp size_t;
// O is a VectorImage AND F has a fixed OuptutSize static constexrp size_t;
template <class F, class T, size_t N> struct NumberOfOutputComponents<F,otb::VectorImage<T>,N>
{
static void Set(const F & f, otb::VectorImage<T> * outputImage, std::array<size_t,N> inNbBands)
......@@ -276,24 +277,21 @@ template <typename C, typename R, typename... T> struct FunctorFilterSuperclassH
using InputHasNeighborhood = std::tuple<typename IsNeighborhood<T>::ValueType...>;
};
// template <class F> class NumberOfOutputDecorator
// {
// public:
// NumberOfOutputDecorator(const F & f, size_t n) : m_Functor(f), m_NumberOfOutputs(n) {}
// template<typename ...T> auto operator()(T...args)
// {
// return m_Functor(args...);
// }
// constexpr size_t OutputSize(...) const
// {
// return m_NumberOfOutputs;
// }
// private:
// F m_Functor;
// size_t m_NumberOfOutputs;
// };
template <typename F> struct NumberOfOutputBandsDecorator : F
{
public:
NumberOfOutputBandsDecorator(const F t, unsigned int nbComp) : F(t), m_NumberOfOutputComponents(nbComp) {}
using F::operator();
constexpr size_t OutputSize(...) const
{
return m_NumberOfOutputComponents;
}
private:
unsigned int m_NumberOfOutputComponents;
};
/** \class FunctorImageFilter
* \brief Implements
......@@ -303,7 +301,10 @@ template <typename C, typename R, typename... T> struct FunctorFilterSuperclassH
* \ingroup IntensityImageFilters Multithreaded
*
* \ingroup OTBImageManipulation
*/
*/
template <typename Functor> auto NewFunctorFilter(const Functor& f, itk::Size<2> radius = {{0,0}});
template <typename Functor> auto NewFunctorFilter(const Functor& f, unsigned int numberOfOuptutBands, itk::Size<2> radius = {{0,0}});
template <class TFunction>
class ITK_EXPORT FunctorImageFilter
: public FunctorFilterSuperclassHelper<TFunction>::FilterType
......@@ -324,7 +325,7 @@ public:
using InputHasNeighborhood = typename FunctorFilterSuperclassHelper<TFunction>::InputHasNeighborhood;
/** Method for creation by passing the filter type. */
static Pointer New(const TFunction& f, itk::Size<2> radius = {{0,0}});
// static Pointer New(const TFunction& f, itk::Size<2> radius = {{0,0}});
/** Run-time type information (and related methods). */
itkTypeMacro(FunctorImageFilter, ImageToImageFilter);
......@@ -362,14 +363,14 @@ void SetFunctor(const FunctorType& functor)
}
private:
friend auto NewFunctorFilter<TFunction>(const TFunction& f, itk::Size<2> radius);
FunctorImageFilter();
FunctorImageFilter(const FunctorType& f, itk::Size<2> radius) : m_Functor(f), m_Radius(radius) {};
FunctorImageFilter(const Self &) ;
void operator =(const Self&) ;
~FunctorImageFilter() {}
protected:
/** FunctorImageFilter can be implemented as a multithreaded filter.
* Therefore, this implementation provides a ThreadedGenerateData() routine
* which is called for each processing thread. The output image data is
......@@ -394,6 +395,23 @@ protected:
itk::Size<2> m_Radius;
};
template <typename Functor> auto NewFunctorFilter(const Functor& f, itk::Size<2> radius)
{
using FilterType = FunctorImageFilter<Functor>;
using PointerType = typename FilterType::Pointer;
PointerType p = new FilterType(f,radius);
return p;
}
template <typename Functor> auto NewFunctorFilter(const Functor& f, unsigned int numberOfOutputBands, itk::Size<2> radius)
{
using FunctorType = NumberOfOutputBandsDecorator<Functor>;
FunctorType decoratedF(f,numberOfOutputBands);
return NewFunctorFilter(decoratedF,radius);
}
}// namespace otb
#ifndef OTB_MANUAL_INSTANTIATION
......
Modules/Filtering/ImageManipulation/include/otbFunctorImageFilter.hxx
+ 0
− 8
View file @ 6eb7f0ca
......@@ -32,14 +32,6 @@ template <class TFunction>
FunctorImageFilter<TFunction>::FunctorImageFilter()
{}
template <class TFunction>
typename FunctorImageFilter<TFunction>::Pointer
FunctorImageFilter<TFunction>::New(const TFunction& f, itk::Size<2> radius)
{
Pointer p = new FunctorImageFilter<TFunction>(f,radius);
return p;
}
template <class TFunction>
void
FunctorImageFilter<TFunction>
......
Modules/Filtering/ImageManipulation/test/otbFunctorImageFilter.cxx
+ 55
− 35
View file @ 6eb7f0ca
......@@ -42,26 +42,6 @@ template <typename O, typename ...T> struct VariadicAdd
}
};
// Redefinition of AddImageFilter, fully variadic
template<typename TOutputImage,
typename ... TInputImage>
using AddImageFilter= otb::FunctorImageFilter<
VariadicAdd<typename TOutputImage::PixelType,
typename TInputImage::PixelType...> >;
// Temptative make_add_filter function to make things even simpler
template<typename O, typename ...T> auto make_add_filter(typename otb::Image<T>::Pointer... imgs)
{
using Filter = AddImageFilter<otb::Image<O>,otb::Image<T>...>;
auto filter = Filter::New(typename Filter::FunctorType{});
filter->SetVInputs(imgs...);
return filter;
}
// helper function to implement next functor (convert a scalar value
// to a VariableLengthVector)
template <typename T> itk::VariableLengthVector<T> toVector(const T & in)
......@@ -170,15 +150,32 @@ template<typename TOut, typename TIn> struct Mean
}
};
// 1 Image with neighborhood of VariableLengthVector -> 1 image with
// VariableLengthVector
// For each channel, returns the maximum value in neighborhood
template<typename T> struct MaxInEachChannel
{
auto operator()(const itk::Neighborhood<itk::VariableLengthVector<T>> & in) const
{
auto out = in.GetCenterValue();
for(auto it = in.Begin(); it!=in.End(); ++it)
{
for(auto band = 0u; band < out.Size();++band)
{
if((*it)[band]>out[band])
out[band] = (*it)[band];
}
}
return out;
}
size_t OutputSize(const std::array<size_t,1> & nbBands) const
{
return nbBands[0];
}
};
// // A lambda
// auto Lambda2 = [](double p)
// {
// itk::VariableLengthVector<double> ret(3);
// ret.Fill(p);
// return ret;
// };
using OImage = typename otb::Image<int>;
using Neig = typename itk::Neighborhood<int>;
......@@ -229,37 +226,60 @@ int otbFunctorImageFilter(int itkNotUsed(argc), char * itkNotUsed(argv) [])
{
return scale*p;
};
auto filterLambda = otb::FunctorImageFilter<decltype(Lambda1)>::New(Lambda1);
auto filterLambda = NewFunctorFilter(Lambda1);
filterLambda->SetVInputs(image);
filterLambda->Update();
// test FunctorImageFilter with a lambda that returns a
// VariableLengthVector
// Converts a neighborhood to a VariableLengthVector
auto Lambda2 = [](const itk::Neighborhood<double> & in)
{
itk::VariableLengthVector<double> out(in.Size());
std::size_t idx{0};
for(auto it = in.Begin(); it!=in.End();++it,++idx)
{
out[idx]=*it;
}
return out;
};
// In this case, we use the helper function which allows to specify
// the number of outputs
auto filterLambda2 = NewFunctorFilter(Lambda2,vimage->GetNumberOfComponentsPerPixel(),{{3,3}});
filterLambda2->SetVInputs(image);
filterLambda2->Update();
// Test FunctorImageFilter with the VariadicConcatenate operator
using ConcatFunctorType = VariadicConcatenate<double, double, itk::VariableLengthVector<double> >;
auto concatenate = otb::FunctorImageFilter<ConcatFunctorType>::New(ConcatFunctorType{});
auto concatenate = NewFunctorFilter(ConcatFunctorType{});
concatenate->SetVInputs(image,vimage);
concatenate->Update();
// Test FunctorImageFilter With VariadicAdd functor
using AddFunctorType = VariadicAdd<double, double, double>;
auto add = otb::FunctorImageFilter<AddFunctorType>::New(AddFunctorType{});
auto add = NewFunctorFilter(AddFunctorType{});
add->SetVInputs(image,image);
add->Update();
// Example of simple helper function to make things easier
auto add2 = make_add_filter<double,double,double>(image,image);
// Test FunctorImageFilter with BandExtraction functor
using ExtractFunctorType = BandExtraction<double,double>;
ExtractFunctorType extractFunctor{1,2};
auto extract = otb::FunctorImageFilter<ExtractFunctorType>::New(extractFunctor);
auto extract = NewFunctorFilter(extractFunctor);
extract->SetVInputs(vimage);
extract->Update();
// Test FunctorImageFilter With Mean functor
using MeanFunctorType = Mean<double,double>;
auto median = otb::FunctorImageFilter<MeanFunctorType>::New(MeanFunctorType{},{{2,2}});
auto median = NewFunctorFilter(MeanFunctorType{},{{2,2}});
median->SetVInputs(image);
median->Update();
// Test FunctorImageFilter with MaxInEachChannel
using MaxInEachChannelType = MaxInEachChannel<double>;
auto maxInEachChannel = NewFunctorFilter(MaxInEachChannelType{},{{3,3}});
maxInEachChannel->SetVInputs(vimage);
maxInEachChannel->Update();
return EXIT_SUCCESS;
}
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment