/*=========================================================================
 *
 *  Copyright NumFOCUS
 *
 *  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
 *
 *         https://www.apache.org/licenses/LICENSE-2.0.txt
 *
 *  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 itkCannyEdgeDetectionImageFilter_h
#define itkCannyEdgeDetectionImageFilter_h

#include "itkConstNeighborhoodIterator.h"
#include "itkDiscreteGaussianImageFilter.h"
#include "itkMultiplyImageFilter.h"
#include "itkMultiThreaderBase.h"
#include "itkDerivativeOperator.h"
#include "itkSparseFieldLayer.h"
#include "itkObjectStore.h"
#include "itkMath.h"

namespace itk
{
template <typename TValue>
class ITK_TEMPLATE_EXPORT ListNode
{
public:
  TValue m_Value;

  ListNode * Next;
  ListNode * Previous;
};

/**
 * \class CannyEdgeDetectionImageFilter
 * \brief This filter is an implementation of a Canny edge detector for
 * scalar-valued images.
 *
 *  Based on John Canny's paper "A Computational Approach
 * to Edge Detection"(IEEE Transactions on Pattern Analysis and Machine
 * Intelligence, Vol. PAMI-8, No.6, November 1986),  there are four major steps
 * used in the edge-detection scheme:
 * (1) Smooth the input image with Gaussian filter.
 * (2) Calculate the second directional derivatives of the smoothed image.
 * (3) Non-Maximum Suppression: the zero-crossings of 2nd derivative are found,
 *     and the sign of third derivative is used to find the correct extrema.
 * (4) The hysteresis thresholding is applied to the gradient magnitude
 *      (multiplied with zero-crossings) of the smoothed image to find and
 *      link edges.
 *
 * \par Inputs and Outputs
 * The input to this filter should be a scalar, real-valued Itk image of
 * arbitrary dimension.  The output should also be a scalar, real-value Itk
 * image of the same dimensionality.
 *
 * \par Parameters
 * There are four parameters for this filter that control the sub-filters used
 * by the algorithm.
 *
 * \par
 * Variance and Maximum error are used in the Gaussian smoothing of the input
 * image.  See  itkDiscreteGaussianImageFilter for information on these
 * parameters.
 *
 * \par
 * Threshold is the lowest allowed value in the output image.  Its data type is
 * the same as the data type of the output image. Any values below the
 * Threshold level will be replaced with the OutsideValue parameter value, whose
 * default is zero.
 *
 * \todo Edge-linking will be added when an itk connected component labeling
 * algorithm is available.
 *
 * \sa DiscreteGaussianImageFilter
 * \sa ZeroCrossingImageFilter
 * \sa ThresholdImageFilter
 * \ingroup ITKImageFeature
 */
template <typename TInputImage, typename TOutputImage>
class ITK_TEMPLATE_EXPORT CannyEdgeDetectionImageFilter : public ImageToImageFilter<TInputImage, TOutputImage>
{
public:
  ITK_DISALLOW_COPY_AND_MOVE(CannyEdgeDetectionImageFilter);

  /** Standard "Self" & Superclass type alias.  */
  using Self = CannyEdgeDetectionImageFilter;
  using Superclass = ImageToImageFilter<TInputImage, TOutputImage>;

  /** Image type alias support   */
  using InputImageType = TInputImage;
  using OutputImageType = TOutputImage;

  /** SmartPointer type alias support  */
  using Pointer = SmartPointer<Self>;
  using ConstPointer = SmartPointer<const Self>;

  /** Define pixel types. */
  using InputImagePixelType = typename TInputImage::PixelType;
  using OutputImagePixelType = typename TOutputImage::PixelType;
  using IndexType = typename TInputImage::IndexType;
  using SizeValueType = typename TInputImage::SizeValueType;

  /** The default boundary condition is used unless overridden
   *in the Evaluate() method. */
  using DefaultBoundaryConditionType = ZeroFluxNeumannBoundaryCondition<OutputImageType>;

  /** The type of data structure that is passed to this function object
   * to evaluate at a pixel that does not lie on a data set boundary.
   */
  using NeighborhoodType = ConstNeighborhoodIterator<OutputImageType, DefaultBoundaryConditionType>;

  using ListNodeType = ListNode<IndexType>;
  using ListNodeStorageType = ObjectStore<ListNodeType>;
  using ListType = SparseFieldLayer<ListNodeType>;
  using ListPointerType = typename ListType::Pointer;

  /** Method for creation through the object factory. */
  itkNewMacro(Self);

  /** Typedef to describe the output image region type. */
  using OutputImageRegionType = typename TOutputImage::RegionType;
  using InputImageRegionType = typename TInputImage::RegionType;

  /** \see LightObject::GetNameOfClass() */
  itkOverrideGetNameOfClassMacro(CannyEdgeDetectionImageFilter);

  /** ImageDimension constant. */
  static constexpr unsigned int ImageDimension = TInputImage::ImageDimension;
  static constexpr unsigned int OutputImageDimension = TOutputImage::ImageDimension;

  /** Typedef of double containers. */
  using ArrayType = FixedArray<double, Self::ImageDimension>;

  /** Set/Get the variance of the Gaussian smoothing filter. */
  itkSetMacro(Variance, ArrayType);
  itkGetConstMacro(Variance, const ArrayType);

  /** Set/Get the maximum error of the Gaussian smoothing kernel in each dimensional
   *  direction. */
  itkSetMacro(MaximumError, ArrayType);
  itkGetConstMacro(MaximumError, const ArrayType);

  /** Set/Get the variance of the Gaussian smoothing filter. */
  void
  SetVariance(const typename ArrayType::ValueType v)
  {
    for (unsigned int i = 0; i < TInputImage::ImageDimension; ++i)
    {
      if (Math::NotExactlyEquals(m_Variance[i], v))
      {
        m_Variance.Fill(v);
        this->Modified();
        break;
      }
    }
  }

  /** Set/Get the MaximumError parameter used by the Gaussian smoothing filter
   *  in this algorithm */
  void
  SetMaximumError(const typename ArrayType::ValueType v)
  {
    for (unsigned int i = 0; i < TInputImage::ImageDimension; ++i)
    {
      if (Math::NotExactlyEquals(m_MaximumError[i], v))
      {
        m_MaximumError.Fill(v);
        this->Modified();
        break;
      }
    }
  }

  /** \brief Set the upper threshold value for detected edges.
   */
  itkSetMacro(UpperThreshold, OutputImagePixelType);
  itkGetConstMacro(UpperThreshold, OutputImagePixelType);

  /** \brief Set the lower threshold value for detected edges.
   */
  itkSetMacro(LowerThreshold, OutputImagePixelType);
  itkGetConstMacro(LowerThreshold, OutputImagePixelType);

  OutputImageType *
  GetNonMaximumSuppressionImage()
  {
    return this->m_MultiplyImageFilter->GetOutput();
  }

#ifdef ITK_USE_CONCEPT_CHECKING
  // Begin concept checking
  itkConceptMacro(InputHasNumericTraitsCheck, (Concept::HasNumericTraits<InputImagePixelType>));
  itkConceptMacro(OutputHasNumericTraitsCheck, (Concept::HasNumericTraits<OutputImagePixelType>));
  itkConceptMacro(SameDimensionCheck, (Concept::SameDimension<ImageDimension, OutputImageDimension>));
  itkConceptMacro(InputIsFloatingPointCheck, (Concept::IsFloatingPoint<InputImagePixelType>));
  itkConceptMacro(OutputIsFloatingPointCheck, (Concept::IsFloatingPoint<OutputImagePixelType>));
  // End concept checking
#endif

protected:
  CannyEdgeDetectionImageFilter();
  void
  PrintSelf(std::ostream & os, Indent indent) const override;

  void
  GenerateData() override;

  using GaussianImageFilterType = DiscreteGaussianImageFilter<InputImageType, OutputImageType>;
  using MultiplyImageFilterType = MultiplyImageFilter<OutputImageType, OutputImageType, OutputImageType>;

private:
  ~CannyEdgeDetectionImageFilter() override = default;

  /** Allocate storage for update buffers used during calculation of multiple steps. */
  void
  AllocateUpdateBuffer();

  /** Implement hysteresis thresholding. */
  void
  HysteresisThresholding();

  /** Edge linking function. */
  void
  FollowEdge(IndexType index, const OutputImageType * multiplyImageFilterOutput);

  /** Calculate the second derivative of the smoothed image, it writes the
   *  result to the update buffer */
  void
  ThreadedCompute2ndDerivative(const OutputImageRegionType & outputRegionForThread);


  /** This method is used to calculate the 2nd derivative for
   * non-boundary pixels. It is called by the ThreadedCompute2ndDerivative
   * method. */
  OutputImagePixelType
  ComputeCannyEdge(const NeighborhoodType & it, void * globalData);

  /** Calculate the gradient of the second derivative of the smoothed image,
   *  it writes the result to m_UpdateBuffer1 */
  void
  ThreadedCompute2ndDerivativePos(const OutputImageRegionType & outputRegionForThread);

  ArrayType m_Variance{};
  ArrayType m_MaximumError{};

  OutputImagePixelType m_UpperThreshold{}; // should be float here?
  OutputImagePixelType m_LowerThreshold{}; // should be float here?

  typename OutputImageType::Pointer m_UpdateBuffer1{};

  /** Gaussian filter to smooth the input image. */
  typename GaussianImageFilterType::Pointer m_GaussianFilter{};

  /** Multiply image filter to multiply with the zero crossings of the second
   *  derivative. */
  typename MultiplyImageFilterType::Pointer m_MultiplyImageFilter{};

  /** Function objects that are used in the inner loops of derivative
   *  calculations. */
  DerivativeOperator<OutputImagePixelType, Self::ImageDimension> m_ComputeCannyEdge1stDerivativeOper{};
  DerivativeOperator<OutputImagePixelType, Self::ImageDimension> m_ComputeCannyEdge2ndDerivativeOper{};

  std::slice m_ComputeCannyEdgeSlice[ImageDimension]{};

  SizeValueType m_Stride[ImageDimension]{};
  SizeValueType m_Center{};

  typename ListNodeStorageType::Pointer m_NodeStore{};
  ListPointerType                       m_NodeList{};

  OutputImageType * m_OutputImage{};
};
} // end of namespace itk

#ifndef ITK_MANUAL_INSTANTIATION
#  include "itkCannyEdgeDetectionImageFilter.hxx"
#endif

#endif