otbDifferenceImageFilter.hxx

Go to the documentation of this file.

/*
 * Copyright (C) 2005-2022 Centre National d'Etudes Spatiales (CNES)
 *
 * This file is part of Orfeo Toolbox
 *
 *     https://www.orfeo-toolbox.org/
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
 
#ifndef otbDifferenceImageFilter_hxx
#define otbDifferenceImageFilter_hxx
 
#include "otbDifferenceImageFilter.h"
 
#include "itkConstNeighborhoodIterator.h"
#include "itkImageRegionIterator.h"
#include "itkNeighborhoodAlgorithm.h"
#include "itkProgressReporter.h"
#include "itkDefaultConvertPixelTraits.h"
 
namespace otb
{
 
//----------------------------------------------------------------------------
template <class TInputImage, class TOutputImage>
DifferenceImageFilter<TInputImage, TOutputImage>::DifferenceImageFilter()
{
  // We require two inputs to execute.
  this->SetNumberOfRequiredInputs(2);
 
  // Set the default DifferenceThreshold.
  m_DifferenceThreshold = itk::NumericTraits<ScalarRealType>::Zero;
 
  // Set the default ToleranceRadius.
  m_ToleranceRadius = 0;
 
  // Initialize statistics about difference image.
  itk::NumericTraits<RealType>::SetLength(m_MeanDifference, itk::DefaultConvertPixelTraits<RealType>::GetNumberOfComponents());
  itk::NumericTraits<AccumulateType>::SetLength(m_TotalDifference, itk::DefaultConvertPixelTraits<RealType>::GetNumberOfComponents());
  m_MeanDifference                = itk::NumericTraits<RealType>::ZeroValue(m_MeanDifference);
  m_TotalDifference               = itk::NumericTraits<AccumulateType>::ZeroValue(m_TotalDifference);
  m_NumberOfPixelsWithDifferences = 0;
  
}
 
//----------------------------------------------------------------------------
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::PrintSelf(std::ostream& os, itk::Indent indent) const
{
  this->Superclass::PrintSelf(os, indent);
  os << indent << "ToleranceRadius: " << m_ToleranceRadius << "\n";
  os << indent << "DifferenceThreshold: " << m_DifferenceThreshold << "\n";
  os << indent << "MeanDifference: " << m_MeanDifference << "\n";
  os << indent << "TotalDifference: " << m_TotalDifference << "\n";
  os << indent << "NumberOfPixelsWithDifferences: " << m_NumberOfPixelsWithDifferences << "\n";
}
 
//----------------------------------------------------------------------------
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::SetValidInput(const InputImageType* validImage)
{
  // The valid image should be input 0.
  this->SetInput(0, validImage);
}
 
//----------------------------------------------------------------------------
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::SetTestInput(const InputImageType* testImage)
{
  // The test image should be input 1.
  this->SetInput(1, testImage);
}
 
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::GenerateOutputInformation()
{
  Superclass::GenerateOutputInformation();
  if (this->GetInput(0)->GetNumberOfComponentsPerPixel() != this->GetInput(1)->GetNumberOfComponentsPerPixel())
  {
    itkExceptionMacro(<< "Image 1 has " << this->GetInput(0)->GetNumberOfComponentsPerPixel() << " bands, whereas image 2 has "
                      << this->GetInput(1)->GetNumberOfComponentsPerPixel());
  }
  this->GetOutput()->SetNumberOfComponentsPerPixel(this->GetInput(0)->GetNumberOfComponentsPerPixel());
}
 
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::Reset()
{
  this->UpdateOutputInformation();
 
  int numberOfThreads = this->GetNumberOfThreads();
 
  itk::NumericTraits<RealType>::SetLength(m_MeanDifference, this->GetInput(0)->GetNumberOfComponentsPerPixel());
  itk::NumericTraits<AccumulateType>::SetLength(m_TotalDifference, this->GetInput(0)->GetNumberOfComponentsPerPixel());
 
  // Initialize statistics about difference image.
  m_MeanDifference                = itk::NumericTraits<RealType>::ZeroValue(m_MeanDifference);
  m_TotalDifference               = itk::NumericTraits<AccumulateType>::ZeroValue(m_TotalDifference);
  m_NumberOfPixelsWithDifferences = 0;
 
  // Resize the thread temporaries
  m_ThreadDifferenceSum.reserve(numberOfThreads);
  m_ThreadNumberOfPixels.SetSize(numberOfThreads);
 
  // Initialize the temporaries
  for (int i = 0; i < numberOfThreads; ++i)
  {
    m_ThreadDifferenceSum.push_back(m_TotalDifference);
  }
  m_ThreadNumberOfPixels.Fill(0);
}
 
//----------------------------------------------------------------------------
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::ThreadedGenerateData(const OutputImageRegionType& threadRegion, itk::ThreadIdType threadId)
{
  typedef itk::ConstNeighborhoodIterator<InputImageType>                           SmartIterator;
  typedef itk::ImageRegionConstIterator<InputImageType>                            InputIterator;
  typedef itk::ImageRegionIterator<OutputImageType>                                OutputIterator;
  typedef itk::NeighborhoodAlgorithm::ImageBoundaryFacesCalculator<InputImageType> FacesCalculator;
  typedef typename FacesCalculator::RadiusType                                     RadiusType;
  typedef typename FacesCalculator::FaceListType                                   FaceListType;
  typedef typename FaceListType::iterator                                          FaceListIterator;
  typedef typename InputImageType::PixelType                                       InputPixelType;
 
  // Prepare standard boundary condition.
  itk::ZeroFluxNeumannBoundaryCondition<InputImageType> nbc;
 
  // Get a pointer to each image.
  const InputImageType* validImage = this->GetInput(0);
  const InputImageType* testImage  = this->GetInput(1);
  OutputImageType*      outputPtr  = this->GetOutput();
 
  // Create a radius of pixels.
  RadiusType radius;
  radius.Fill(std::max(0, m_ToleranceRadius));
 
  // Find the data-set boundary faces.
  FacesCalculator boundaryCalculator;
  FaceListType    faceList = boundaryCalculator(testImage, threadRegion, radius);
 
  // Support progress methods/callbacks.
  itk::ProgressReporter progress(this, threadId, threadRegion.GetNumberOfPixels());
 
  AccumulateType threadDifferenceSum;
  itk::NumericTraits<AccumulateType>::SetLength(threadDifferenceSum, this->GetInput(0)->GetNumberOfComponentsPerPixel()); // @post: threadDifferenceSum=={0...}
  unsigned long threadNumberOfPixels = 0;
 
  // Process the internal face and each of the boundary faces.
  for (FaceListIterator face = faceList.begin(); face != faceList.end(); ++face)
  {
    SmartIterator  test(radius, testImage, *face); // Iterate over test image.
    InputIterator  valid(validImage, *face);       // Iterate over valid image.
    OutputIterator out(outputPtr, *face);          // Iterate over output image.
    test.OverrideBoundaryCondition(&nbc);
 
    // Extract a typical pixel in order to know the exact size of the
    // pixel, the typical 0, and the typical max
    valid.GoToBegin();
    InputPixelType t = valid.Get();
 
    const auto pixel_size = itk::NumericTraits<InputPixelType>::GetLength(t);
    const auto out_max    = itk::NumericTraits<OutputPixelType>::max(t);
    const auto zero       = itk::NumericTraits<OutputPixelType>::ZeroValue(t);
 
    OutputPixelType minimumDifference = out_max;
 
    for (valid.GoToBegin(), test.GoToBegin(), out.GoToBegin(); !valid.IsAtEnd(); ++valid, ++test, ++out)
    {
      // Get the current valid pixel.
      t = valid.Get();
 
      // Find the closest-valued pixel in the neighborhood of the test
      // image.
      minimumDifference             = out_max; // reset by assignment, avoid allocation in VLV case
      unsigned int neighborhoodSize = test.Size();
      for (unsigned int i = 0; i < neighborhoodSize; ++i)
      {
 
        for (unsigned int j = 0; j < pixel_size; ++j)
        {
          // Use the RealType for the difference to make sure we get
          // the sign.
          // Work on component independently in order to avoid
          // allocation of VLV pixels
          ScalarRealType d = static_cast<ScalarRealType>(itk::DefaultConvertPixelTraits<InputPixelType>::GetNthComponent(j, t) -
                                                         itk::DefaultConvertPixelTraits<InputPixelType>::GetNthComponent(j, test.GetPixel(i)));
          d                = std::abs(d);
          ScalarRealType m = static_cast<ScalarRealType>(itk::DefaultConvertPixelTraits<OutputPixelType>::GetNthComponent(j, minimumDifference));
          if (d < m)
          {
            itk::DefaultConvertPixelTraits<OutputPixelType>::SetNthComponent(j, minimumDifference, d);
            //             std::cout << std::setprecision(16) << minimumDifference[j] << std::endl;
            //             std::cout << std::setprecision(16) << t << std::endl;
            //             std::cout << std::setprecision(16) << test.GetPixel(i) << std::endl;
            //             std::cout << "----------------------" << std::endl;
          }
        }
      }
 
      // for complex and vector type. FIXME: module might be better
      //        ScalarRealType tMax=std::abs(t[0]);
      ScalarRealType tMax = 0.01; // Avoiding the 0 case for neighborhood computing
      // NB: still more restrictive than before for small values.
      for (unsigned int j = 0; j < pixel_size; ++j)
      {
        ScalarRealType tc = static_cast<ScalarRealType>(itk::DefaultConvertPixelTraits<InputPixelType>::GetNthComponent(j, t));
        tMax              = std::max(std::abs(tc), tMax);
      }
 
      // Check if difference is above threshold
      // the threshold is interpreted as relative to the value
      bool isDifferent = false;
 
      for (unsigned int j = 0; j < pixel_size; ++j)
      {
        ScalarRealType m = static_cast<ScalarRealType>(itk::DefaultConvertPixelTraits<OutputPixelType>::GetNthComponent(j, minimumDifference));
        if (m > m_DifferenceThreshold * tMax)
        {
          //           std::cout << std::setprecision(16) << minimumDifference[j] << std::endl;
          isDifferent = true;
          break;
        }
      }
 
      if (isDifferent)
      {
        // Store the minimum difference value in the output image.
        out.Set(minimumDifference);
 
        // Update local difference image statistics.
        threadDifferenceSum += minimumDifference;
        threadNumberOfPixels++;
      }
      else
      {
        // Difference is below threshold.
        out.Set(zero);
      }
 
      // Update progress.
      progress.CompletedPixel();
    }
  }
 
  // Update global difference image statistics.
  m_ThreadDifferenceSum[threadId]  += threadDifferenceSum;
  m_ThreadNumberOfPixels[threadId] += threadNumberOfPixels;
}
 
template <class TInputImage, class TOutputImage>
void DifferenceImageFilter<TInputImage, TOutputImage>::Synthetize()
{
  // Set statistics about difference image.
  int numberOfThreads = this->GetNumberOfThreads();
  for (int i = 0; i < numberOfThreads; ++i)
  {
    m_TotalDifference += m_ThreadDifferenceSum[i];
    m_NumberOfPixelsWithDifferences += m_ThreadNumberOfPixels[i];
  }
 
  // Get the total number of pixels processed in the region.
  // This is different from the m_TotalNumberOfPixels which
  // is the number of pixels that actually have differences
  // above the intensity threshold.
  OutputImageRegionType region         = this->GetOutput()->GetRequestedRegion();
  unsigned int          numberOfPixels = region.GetNumberOfPixels();
 
  // Calculate the mean difference.
  m_MeanDifference = m_TotalDifference / numberOfPixels;
}
 
 
 
} // end namespace otb
 
#endif