/*=========================================================================
 *
 *  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 itkScalarImageToTextureFeaturesFilter_h
#define itkScalarImageToTextureFeaturesFilter_h

#include "itkDataObjectDecorator.h"

#include "itkHistogramToTextureFeaturesFilter.h"
#include "itkScalarImageToCooccurrenceMatrixFilter.h"

namespace itk
{
namespace Statistics
{
/** \class ScalarImageToTextureFeaturesFilter
 *  \brief This class computes texture descriptions from an image.
 *
 * This class computes features that summarize the texture of a given image.
 * The texture features are computed a la Haralick, and have proven to be useful
 * in image classification for biological and medical imaging.
 * This class computes the texture features of an image (optionally in a
 * masked region), averaged across several spatial directions so that they are
 * invariant to rotation.
 *
 * By default, texture features are computed for each spatial
 * direction and then averaged afterward, so it is possible to access the standard
 * deviations of the texture features. These values give a clue as to texture
 * anisotropy. However, doing this is much more work, because it involved computing
 * one GLCM for each offset given. To compute a single GLCM using the first offset ,
 * call FastCalculationsOn(). If this is called, then the texture standard deviations
 * will not be computed (and will be set to zero), but texture computation will
 * be much faster.
 *
 * This class is templated over the input image type.
 *
 * Template Parameters:
 * The image type, and the type of histogram frequency container. If you are using
 * a large number of bins per axis, a sparse frequency container may be advisable.
 * The default is to use a dense frequency container.
 *
 * Inputs and parameters:
 * -# An image
 * -# A mask defining the region over which texture features will be
 *    calculated. (Optional)
 * -# The pixel value that defines the "inside" of the mask. (Optional, defaults
 *    to 1 if a mask is set.)
 * -# The set of features to be calculated. These features are defined
 *    in the GreyLevelCooccurrenceMatrixTextureCoefficientsCalculator class. (Optional,
 *    defaults to {Energy, Entropy, InverseDifferenceMoment, Inertia, ClusterShade,
 *    ClusterProminence}, as in Conners, Trivedi and Harlow.)
 * -# The number of intensity bins. (Optional, defaults to 256.)
 * -# The set of directions (offsets) to average across. (Optional, defaults to
 *    {(-1, 0), (-1, -1), (0, -1), (1, -1)} for 2D images and scales analogously for ND
 *    images.)
 * -# The pixel intensity range over which the features will be calculated.
 *    (Optional, defaults to the full dynamic range of the pixel type.)
 *
 * In general, the default parameter values should be sufficient.
 *
 * Outputs:
 * (1) The average value of each feature.
 * (2) The standard deviation in the values of each feature.
 *
 * Web reference:
 * http://www.fp.ucalgary.ca/mhallbey/tutorial.htm
 *
 * Print references:
 * Haralick, R.M., K. Shanmugam and I. Dinstein. 1973.  Textural Features for
 * Image Classification. IEEE Transactions on Systems, Man and Cybernetics.
 * SMC-3(6):610-620.
 *
 * Haralick, R.M. 1979. Statistical and Structural Approaches to Texture.
 * Proceedings of the IEEE, 67:786-804.
 *
 * R.W. Conners and C.A. Harlow. A Theoretical Comparison of Texture Algorithms.
 * IEEE Transactions on Pattern Analysis and Machine Intelligence,  2:204-222, 1980.
 *
 * R.W. Conners, M.M. Trivedi, and C.A. Harlow. Segmentation of a High-Resolution
 * Urban Scene using Texture  Operators. Computer Vision, Graphics and Image
 * Processing, 25:273-310,  1984.
 *
 * \sa ScalarImageToCooccurrenceMatrixFilter
 * \sa HistogramToTextureFeaturesFilter
 *
 * Author: Zachary Pincus
 * \ingroup ITKStatistics
 *
 * \sphinx
 * \sphinxexample{Numerics/Statistics/ComputeTextureFeatures,Compute Texture Features}
 * \endsphinx
 */

template <typename TImageType,
          typename THistogramFrequencyContainer = DenseFrequencyContainer2,
          typename TMaskImageType = TImageType>
class ITK_TEMPLATE_EXPORT ScalarImageToTextureFeaturesFilter : public ProcessObject
{
public:
  /** Standard type alias */
  using Self = ScalarImageToTextureFeaturesFilter;
  using Superclass = ProcessObject;
  using Pointer = SmartPointer<Self>;
  using ConstPointer = SmartPointer<const Self>;

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

  /** standard New() method support */
  itkNewMacro(Self);

  using FrequencyContainerType = THistogramFrequencyContainer;
  using ImageType = TImageType;
  using ImagePointer = typename ImageType::Pointer;
  using MaskImageType = TMaskImageType;
  using MaskPointer = typename MaskImageType::Pointer;
  using MaskPixelType = typename MaskImageType::PixelType;

  using PixelType = typename ImageType::PixelType;
  using OffsetType = typename ImageType::OffsetType;
  using OffsetVector = VectorContainer<unsigned char, OffsetType>;
  using OffsetVectorPointer = typename OffsetVector::Pointer;
  using OffsetVectorConstPointer = typename OffsetVector::ConstPointer;

  using CooccurrenceMatrixFilterType =
    ScalarImageToCooccurrenceMatrixFilter<ImageType, FrequencyContainerType, MaskImageType>;

  using HistogramType = typename CooccurrenceMatrixFilterType::HistogramType;
  using TextureFeaturesFilterType = HistogramToTextureFeaturesFilter<HistogramType>;

  // More work needs to be done to fix wrapping
  // using TextureFeatureName = itk::Statistics::TextureFeatureEnum;
  using TextureFeatureName = uint8_t;
  using FeatureNameVector = VectorContainer<unsigned char, TextureFeatureName>;

  using FeatureNameVectorPointer = typename FeatureNameVector::Pointer;
  using FeatureNameVectorConstPointer = typename FeatureNameVector::ConstPointer;
  using FeatureValueVector = VectorContainer<unsigned char, double>;
  using FeatureValueVectorPointer = typename FeatureValueVector::Pointer;

  /** Smart Pointer type to a DataObject. */
  using DataObjectPointer = DataObject::Pointer;

  /** Type of DataObjects used for scalar outputs */
  using FeatureValueVectorDataObjectType = DataObjectDecorator<FeatureValueVector>;

  const FeatureValueVectorDataObjectType *
  GetFeatureMeansOutput() const;

  const FeatureValueVectorDataObjectType *
  GetFeatureStandardDeviationsOutput() const;

  /** Connects the input image for which the features are going to be computed
   */
  using Superclass::SetInput;
  void
  SetInput(const ImageType *);

  const ImageType *
  GetInput() const;

  /** Return the feature means and deviations.  */
  itkGetConstReferenceObjectMacro(FeatureMeans, FeatureValueVector);
  itkGetConstReferenceObjectMacro(FeatureStandardDeviations, FeatureValueVector);

  /** Set the desired feature set. Optional, for default value see above. */
  itkSetConstObjectMacro(RequestedFeatures, FeatureNameVector);
  itkGetConstObjectMacro(RequestedFeatures, FeatureNameVector);

  /** Set the  offsets over which the co-occurrence pairs will be computed.
      Optional; for default value see above. */
  itkSetConstObjectMacro(Offsets, OffsetVector);
  itkGetConstObjectMacro(Offsets, OffsetVector);

  /** Set number of histogram bins along each axis.
      Optional; for default value see above. */
  void
  SetNumberOfBinsPerAxis(unsigned int);

  /** Set the min and max (inclusive) pixel value that will be used for
      feature calculations. Optional; for default value see above. */
  void
  SetPixelValueMinMax(PixelType min, PixelType max);

  /** Connects the mask image for which the histogram is going to be computed.
      Optional; for default value see above. */
  void
  SetMaskImage(const MaskImageType *);

  const MaskImageType *
  GetMaskImage() const;

  /** Set the pixel value of the mask that should be considered "inside" the
      object. Optional; for default value see above. */
  void
  SetInsidePixelValue(MaskPixelType insidePixelValue);

  itkGetConstMacro(FastCalculations, bool);
  itkSetMacro(FastCalculations, bool);
  itkBooleanMacro(FastCalculations);

protected:
  ScalarImageToTextureFeaturesFilter();
  ~ScalarImageToTextureFeaturesFilter() override = default;
  void
  PrintSelf(std::ostream & os, Indent indent) const override;

  void
  FastCompute();

  void
  FullCompute();

  /** This method causes the filter to generate its output. */
  void
  GenerateData() override;

  /** Make a DataObject to be used for output output. */
  using DataObjectPointerArraySizeType = ProcessObject::DataObjectPointerArraySizeType;
  using Superclass::MakeOutput;
  DataObjectPointer MakeOutput(DataObjectPointerArraySizeType) override;

private:
  typename CooccurrenceMatrixFilterType::Pointer m_GLCMGenerator{};

  typename TextureFeaturesFilterType::Pointer m_GLCMCalculator{};

  FeatureValueVectorPointer     m_FeatureMeans{};
  FeatureValueVectorPointer     m_FeatureStandardDeviations{};
  FeatureNameVectorConstPointer m_RequestedFeatures{};
  OffsetVectorConstPointer      m_Offsets{};
  bool                          m_FastCalculations{};
};
} // end of namespace Statistics
} // end of namespace itk

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

#endif