Imaging/Color/vtkImageQuantizeRGBToIndex.cxx

/*=========================================================================

  Program:   Visualization Toolkit
  Module:    vtkImageQuantizeRGBToIndex.cxx

  Copyright (c) Ken Martin, Will Schroeder, Bill Lorensen
  All rights reserved.
  See Copyright.txt or http://www.kitware.com/Copyright.htm for details.

     This software is distributed WITHOUT ANY WARRANTY; without even
     the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
     PURPOSE.  See the above copyright notice for more information.

=========================================================================*/
#include "vtkImageQuantizeRGBToIndex.h"

#include "vtkImageData.h"
#include "vtkInformation.h"
#include "vtkInformationVector.h"
#include "vtkLookupTable.h"
#include "vtkObjectFactory.h"
#include "vtkStreamingDemandDrivenPipeline.h"
#include "vtkTimerLog.h"

#include "vtkExtractVOI.h"
#include "vtkImageLuminance.h"
#include "vtkSmartPointer.h"

#include <cmath>

vtkStandardNewMacro(vtkImageQuantizeRGBToIndex);

class vtkColorQuantizeNode
{
public:
  vtkColorQuantizeNode()
  {
    this->Axis = -1;
    this->SplitPoint = -1;
    this->Index = -1;
    this->Child1 = nullptr;
    this->Child2 = nullptr;
    this->StdDev[0] = this->StdDev[1] = this->StdDev[2] = 0.0;
    this->Histogram[0] = this->Histogram[1] = this->Histogram[2] = nullptr;
    this->Image = nullptr;
    this->Bounds[0] = 0;
    this->Bounds[1] = 256;
    this->Bounds[2] = 0;
    this->Bounds[3] = 256;
    this->Bounds[4] = 0;
    this->Bounds[5] = 256;
  }

  ~vtkColorQuantizeNode()
  {
    delete[] this->Histogram[0];
    delete[] this->Histogram[1];
    delete[] this->Histogram[2];
    delete this->Child1;
    delete this->Child2;
  }

  void SetImageExtent(int v[6]) { memcpy(this->ImageExtent, v, 6 * sizeof(int)); }

  void SetImageIncrement(vtkIdType v[3]) { memcpy(this->ImageIncrement, v, 3 * sizeof(vtkIdType)); }

  void SetImageType(double type) { this->ImageType = static_cast<int>(type); }

  void SetImage(void* image) { this->Image = image; }

  void SetAxis(int v) { this->Axis = v; }
  int GetAxis() { return this->Axis; }

  void SetSplitPoint(int v) { this->SplitPoint = v; }
  int GetSplitPoint() { return this->SplitPoint; }

  int* GetBounds() { return this->Bounds; }
  void SetBounds(int v[6]) { memcpy(this->Bounds, v, 6 * sizeof(int)); }

  void SetIndex(int v) { this->Index = v; }
  int GetIndex() { return this->Index; }

  double GetStdDev(int axis) { return this->StdDev[axis]; }
  void ComputeStdDev();

  int GetCount() { return this->Count; }

  double GetMean(int axis) { return this->Mean[axis]; }
  void GetMean(double c[3])
  {
    c[0] = this->Mean[0];
    c[1] = this->Mean[1];
    c[2] = this->Mean[2];
  }

  void Divide(int axis, int nextIndex);

  void SetChild1(vtkColorQuantizeNode* n) { this->Child1 = n; }
  vtkColorQuantizeNode* GetChild1() { return this->Child1; }

  void SetChild2(vtkColorQuantizeNode* n) { this->Child2 = n; }
  vtkColorQuantizeNode* GetChild2() { return this->Child2; }

  int GetIndex(int c[3])
  {
    if (this->Index >= 0)
    {
      return this->Index;
    }
    if (c[this->Axis] > this->SplitPoint)
    {
      return this->Child2->GetIndex(c);
    }
    return this->Child1->GetIndex(c);
  }

  void GetAverageColor(int c[3])
  {
    if (this->AverageCount)
    {
      c[0] = static_cast<int>(this->AverageColor[0] / this->AverageCount);
      c[1] = static_cast<int>(this->AverageColor[1] / this->AverageCount);
      c[2] = static_cast<int>(this->AverageColor[2] / this->AverageCount);
    }
  }

  void StartColorAveraging()
  {
    if (this->Child1)
    {
      this->Child1->StartColorAveraging();
      this->Child2->StartColorAveraging();
    }
    else
    {
      this->AverageCount = 0;
      this->AverageColor[0] = 0.0;
      this->AverageColor[1] = 0.0;
      this->AverageColor[2] = 0.0;
    }
  }

  void AddColor(int c[3])
  {
    this->AverageCount++;
    this->AverageColor[0] += c[0];
    this->AverageColor[1] += c[1];
    this->AverageColor[2] += c[2];
  }

protected:
  int Axis;
  int SplitPoint;
  int Bounds[6];
  int Index;
  double StdDev[3];
  double Median[3];
  double Mean[3];
  int Count;
  int AverageCount;
  double AverageColor[3];
  vtkIdType ImageIncrement[3];
  int ImageExtent[6];
  int ImageType;
  void* Image;
  int* Histogram[3];
  vtkColorQuantizeNode *Child1, *Child2;
};

template <class T>
void vtkImageQuantizeRGBToIndexHistogram(
  T* inPtr, int extent[6], vtkIdType inIncrement[3], int type, int bounds[6], int* histogram[3])
{
  T *rgbPtr, v[3];
  int x, y, z, c;
  int value[3];
  int max[3];

  max[0] = bounds[1] - bounds[0] + 1;
  max[1] = bounds[3] - bounds[2] + 1;
  max[2] = bounds[5] - bounds[4] + 1;

  for (c = 0; c < 3; c++)
  {
    for (x = 0; x < max[c]; x++)
    {
      histogram[c][x] = 0;
    }
  }

  // Generate the histogram
  rgbPtr = inPtr;
  for (z = extent[4]; z <= extent[5]; z++)
  {
    for (y = extent[2]; y <= extent[3]; y++)
    {
      for (x = extent[0]; x <= extent[1]; x++)
      {
        if (type == VTK_UNSIGNED_CHAR)
        {
          v[0] = *(rgbPtr++) - bounds[0];
          v[1] = *(rgbPtr++) - bounds[2];
          v[2] = *(rgbPtr++) - bounds[4];
          if (static_cast<int>(v[0]) < max[0] && static_cast<int>(v[1]) < max[1] &&
            static_cast<int>(v[2]) < max[2])
          {
            histogram[0][static_cast<unsigned char>(v[0])]++;
            histogram[1][static_cast<unsigned char>(v[1])]++;
            histogram[2][static_cast<unsigned char>(v[2])]++;
          }
        }
        else if (type == VTK_UNSIGNED_SHORT)
        {
          v[0] = ((static_cast<unsigned short>(*(rgbPtr++))) >> 8) - bounds[0];
          v[1] = ((static_cast<unsigned short>(*(rgbPtr++))) >> 8) - bounds[2];
          v[2] = ((static_cast<unsigned short>(*(rgbPtr++))) >> 8) - bounds[4];
          if (static_cast<int>(v[0]) < max[0] && static_cast<int>(v[1]) < max[1] &&
            static_cast<int>(v[2]) < max[2])
          {
            histogram[0][static_cast<unsigned short>(v[0])]++;
            histogram[1][static_cast<unsigned short>(v[1])]++;
            histogram[2][static_cast<unsigned short>(v[2])]++;
          }
        }
        else
        {
          value[0] = static_cast<int>(*(rgbPtr++) * 255.5) - bounds[0];
          value[1] = static_cast<int>(*(rgbPtr++) * 255.5) - bounds[2];
          value[2] = static_cast<int>(*(rgbPtr++) * 255.5) - bounds[4];
          if (static_cast<int>(value[0]) < max[0] && static_cast<int>(value[1]) < max[1] &&
            static_cast<int>(value[2]) < max[2])
          {
            histogram[0][value[0]]++;
            histogram[1][value[1]]++;
            histogram[2][value[2]]++;
          }
        }
        rgbPtr += inIncrement[0];
      }
      rgbPtr += inIncrement[1];
    }
    rgbPtr += inIncrement[2];
  }
}

// This templated function executes the filter for supported types of data.
template <class T>
void vtkImageQuantizeRGBToIndexExecute(
  vtkImageQuantizeRGBToIndex* self, vtkImageData* inData, vtkImageData* outData)
{
  int extent[6];
  vtkIdType inIncrement[3], outIncrement[3];
  T* rgbPtr;
  unsigned short* indexPtr;
  int x, y, z, c;
  int type;
  vtkColorQuantizeNode *root, *tmp;
  vtkColorQuantizeNode* leafNodes[65536];
  int numLeafNodes;
  int maxdevAxis = 0, maxdevLeafNode = 0;
  double maxdev, dev;
  int leaf, axis;
  int cannotDivideFurther;
  vtkLookupTable* lut;
  double color[4];
  int rgb[3];
  vtkTimerLog* timer;
  vtkIdType totalCount;
  double weight;

  timer = vtkTimerLog::New();
  timer->StartTimer();
  type = self->GetInputType();

  // need extent to get increments.
  // in and out extents are the same
  inData->GetExtent(extent);

  inData->GetContinuousIncrements(extent, inIncrement[0], inIncrement[1], inIncrement[2]);
  outData->GetContinuousIncrements(extent, outIncrement[0], outIncrement[1], outIncrement[2]);

  timer->StopTimer();

  self->SetInitializeExecuteTime(timer->GetElapsedTime());
  timer->StartTimer();

  vtkImageData* image_sample = inData;
  auto sampler =
    vtkSmartPointer<vtkExtractVOI>::New(); // outside of scope to keep samples image alive
  if (self->GetSamplingRate()[0] > 1 || self->GetSamplingRate()[1] > 1 ||
    self->GetSamplingRate()[2] > 1)
  {
    // The idea behind this sampling is to build the octree
    // with mean colors using a subset of the image data, potentially
    // allowing a significant speedup.
    sampler->SetInputData(inData);
    sampler->SetVOI(inData->GetExtent());
    sampler->SetSampleRate(self->GetSamplingRate());
    sampler->Update();
    image_sample = sampler->GetOutput();
  }

  auto inPtrSampled = static_cast<T*>(image_sample->GetScalarPointer());
  auto inPtr = static_cast<T*>(inData->GetScalarPointer());
  auto outPtr = static_cast<unsigned short*>(outData->GetScalarPointer());

  int extentSampled[6];
  vtkIdType incSampled[3];
  image_sample->GetExtent(extentSampled);
  image_sample->GetContinuousIncrements(extentSampled, incSampled[0], incSampled[1], incSampled[2]);

  // Build the tree
  // Create the root node - it is our only leaf node
  root = new vtkColorQuantizeNode;
  root->SetIndex(0);
  root->SetImageExtent(extentSampled);
  root->SetImageIncrement(incSampled);
  root->SetImageType(type);
  root->SetImage(inPtrSampled);
  root->ComputeStdDev();
  leafNodes[0] = root;
  numLeafNodes = 1;

  cannotDivideFurther = 0;

  totalCount = extentSampled[1] - extentSampled[0] + 1;
  totalCount *= extentSampled[3] - extentSampled[2] + 1;
  totalCount *= extentSampled[5] - extentSampled[4] + 1;

  // Loop until we've added enough leaf nodes or we can't add any more
  // Here we are using sampled image (VOI).
  while (numLeafNodes < self->GetNumberOfColors() && !cannotDivideFurther)
  {
    // Find leaf node / axis with maximum deviation
    maxdev = 0.0;
    for (leaf = 0; leaf < numLeafNodes; leaf++)
    {
      for (axis = 0; axis < 3; axis++)
      {
        dev = leafNodes[leaf]->GetStdDev(axis);
        weight = static_cast<double>(leafNodes[leaf]->GetCount()) / static_cast<double>(totalCount);
        dev *= weight;
        if (dev > maxdev)
        {
          maxdevAxis = axis;
          maxdevLeafNode = leaf;
          maxdev = dev;
        }
      }
    }
    if (maxdev == 0.0)
    {
      cannotDivideFurther = 1;
    }
    else
    {
      leafNodes[maxdevLeafNode]->Divide(maxdevAxis, numLeafNodes);
      leafNodes[numLeafNodes] = leafNodes[maxdevLeafNode]->GetChild1();
      leafNodes[maxdevLeafNode] = leafNodes[maxdevLeafNode]->GetChild2();
      numLeafNodes++;
    }

    self->UpdateProgress(0.6667 * numLeafNodes / self->GetNumberOfColors());
  }

  timer->StopTimer();
  self->SetBuildTreeExecuteTime(timer->GetElapsedTime());
  timer->StartTimer();

  root->StartColorAveraging();

  // Sort color indices by luminance
  // The "index-image" viewed as a greyscale image, is usually quite
  // arbitrary, accentuating contrast where none can be perceived in
  // the original color image.
  // To make the index image more meaningful, we sort the mean colors
  // by luminance (mapping the indices accordingly).
  if (self->GetSortIndexByLuminance())
  {
    std::vector<std::pair<double, int>> luminance_index_map;
    double RGB[3];
    for (leaf = 0; leaf < numLeafNodes; leaf++)
    {
      leafNodes[leaf]->GetMean(RGB);
      double luminance = 0.30 * RGB[0] + 0.59 * RGB[1] + 0.11 * RGB[2];
      int index = leafNodes[leaf]->GetIndex();
      luminance_index_map.emplace_back(luminance, index);
    }
    std::sort(luminance_index_map.begin(), luminance_index_map.end());
    std::vector<int> index_map(luminance_index_map.size());
    for (size_t i = 0; i < luminance_index_map.size(); ++i)
    {
      index_map.at(luminance_index_map[i].second) = static_cast<int>(i);
    }
    for (leaf = 0; leaf < numLeafNodes; leaf++)
    {
      int index = leafNodes[leaf]->GetIndex();
      leafNodes[leaf]->SetIndex(index_map.at(index));
    }
  }

  // Fill in the indices in the output image
  indexPtr = outPtr;
  rgbPtr = inPtr;
  for (z = extent[4]; z <= extent[5]; z++)
  {
    for (y = extent[2]; !self->AbortExecute && y <= extent[3]; y++)
    {
      for (x = extent[0]; x <= extent[1]; x++)
      {
        for (c = 0; c < 3; c++)
        {
          if (type == VTK_UNSIGNED_CHAR)
          {
            rgb[c] = static_cast<int>(*rgbPtr);
          }
          else if (type == VTK_UNSIGNED_SHORT)
          {
            rgb[c] = (static_cast<unsigned short>(*rgbPtr)) >> 8;
          }
          else
          {
            rgb[c] = static_cast<int>(*rgbPtr * 255.5);
          }
          rgbPtr++;
        }
        tmp = root;
        while (true)
        {
          if (tmp->GetIndex() != -1)
          {
            *indexPtr = tmp->GetIndex();
            break;
          }
          if (rgb[tmp->GetAxis()] > tmp->GetSplitPoint())
          {
            tmp = tmp->GetChild2();
          }
          else
          {
            tmp = tmp->GetChild1();
          }
        }
        tmp->AddColor(rgb);
        indexPtr++;

        rgbPtr += inIncrement[0];
        indexPtr += outIncrement[0];
      }
      rgbPtr += inIncrement[1];
      indexPtr += outIncrement[1];
    }
    rgbPtr += inIncrement[2];
    indexPtr += outIncrement[2];
  }

  self->UpdateProgress(0.90);

  // Fill in the lookup table
  lut = self->GetLookupTable();
  lut->SetNumberOfTableValues(numLeafNodes);
  lut->SetNumberOfColors(numLeafNodes);
  lut->SetTableRange(0, numLeafNodes - 1);
  color[3] = 1.0;
  for (leaf = 0; leaf < numLeafNodes; leaf++)
  {
    leafNodes[leaf]->GetAverageColor(rgb);
    color[0] = rgb[0] / 255.0;
    color[1] = rgb[1] / 255.0;
    color[2] = rgb[2] / 255.0;
    lut->SetTableValue(leafNodes[leaf]->GetIndex(), color);
  }

  timer->StopTimer();
  self->SetLookupIndexExecuteTime(timer->GetElapsedTime());
  timer->Delete();

  delete root;
}

void vtkColorQuantizeNode::ComputeStdDev()
{
  int i, j;
  double mean;
  int count = 0;
  int medianCount;

  // Create space for histogram
  this->Histogram[0] = new int[this->Bounds[1] - this->Bounds[0] + 1];
  this->Histogram[1] = new int[this->Bounds[3] - this->Bounds[2] + 1];
  this->Histogram[2] = new int[this->Bounds[5] - this->Bounds[4] + 1];

  // Create histogram
  switch (this->ImageType)
  {
    vtkTemplateMacro(vtkImageQuantizeRGBToIndexHistogram(static_cast<VTK_TT*>(this->Image),
      this->ImageExtent, this->ImageIncrement, this->ImageType, this->Bounds, this->Histogram));
  }

  // Compute for r, g, and b
  for (i = 0; i < 3; i++)
  {
    // Compute the mean
    mean = 0;
    count = 0;
    for (j = 0; j <= (this->Bounds[i * 2 + 1] - this->Bounds[i * 2]); j++)
    {
      count += this->Histogram[i][j];
      mean += this->Histogram[i][j] * (j + this->Bounds[i * 2]);
    }
    if (count > 0)
    {
      mean /= static_cast<double>(count);
    }
    else
    {
      mean = 0;
    }
    this->Mean[i] = mean;

    // Must have some minimum distance to subdivide - if we
    // are below this distance limit, don't compute a
    // standard deviation since we don't want to subdivide this
    // node along this axis. Set the deviation to 0.0 and continue.
    if (this->Bounds[i * 2 + 1] == this->Bounds[i * 2])
    {
      this->StdDev[i] = 0.0;
      continue;
    }

    // Where is the median?
    medianCount = count / 2;

    // Initialize the median to unset
    this->Median[i] = -1;

    // Compute the standard deviation and the location of the median
    this->StdDev[i] = 0;
    count = 0;
    for (j = 0; j <= (this->Bounds[i * 2 + 1] - this->Bounds[i * 2]); j++)
    {
      count += this->Histogram[i][j];
      this->StdDev[i] += static_cast<double>(this->Histogram[i][j]) *
        (static_cast<double>(j) + this->Bounds[i * 2] - mean) *
        (static_cast<double>(j) + this->Bounds[i * 2] - mean);
      if (this->Median[i] == -1 && count > medianCount)
      {
        this->Median[i] = j + this->Bounds[i * 2];
      }
    }

    // If our median is at the upper bound, bump down by one. This will
    // help in the cases where we have a distance of 2 in this dimension,
    // and just over half the entries are in the second bucket. We
    // still want to divide - the division needs to be at the first
    // bucket.
    if (this->Median[i] == this->Bounds[i * 2 + 1])
    {
      this->Median[i]--;
    }

    // Do the final division and square root to get the standard deviation
    if (count > 0)
    {
      this->StdDev[i] /= static_cast<double>(count);
    }
    else
    {
      this->StdDev[i] = 0;
    }

    this->StdDev[i] = sqrt(this->StdDev[i]);
  }

  // Should all be the same - just take the last one
  this->Count = count;
}

void vtkColorQuantizeNode::Divide(int axis, int nextIndex)
{
  int newBounds[6];

  this->Child1 = new vtkColorQuantizeNode;
  this->Child2 = new vtkColorQuantizeNode;

  memcpy(newBounds, this->Bounds, 6 * sizeof(int));

  newBounds[axis * 2 + 1] = static_cast<int>(this->Median[axis]);
  this->Child1->SetBounds(newBounds);

  newBounds[axis * 2] = static_cast<int>(this->Median[axis] + 1);
  newBounds[axis * 2 + 1] = static_cast<int>(this->Bounds[axis * 2 + 1]);
  this->Child2->SetBounds(newBounds);

  this->SplitPoint = static_cast<int>(this->Median[axis]);
  this->Axis = axis;

  this->Child1->SetIndex(this->Index);
  this->Child2->SetIndex(nextIndex);
  this->Index = -1;

  delete[] this->Histogram[0];
  delete[] this->Histogram[1];
  delete[] this->Histogram[2];

  this->Histogram[0] = nullptr;
  this->Histogram[1] = nullptr;
  this->Histogram[2] = nullptr;

  this->Child1->SetImageExtent(this->ImageExtent);
  this->Child1->SetImageIncrement(this->ImageIncrement);
  this->Child1->SetImageType(this->ImageType);
  this->Child1->SetImage(this->Image);

  this->Child2->SetImageExtent(this->ImageExtent);
  this->Child2->SetImageIncrement(this->ImageIncrement);
  this->Child2->SetImageType(this->ImageType);
  this->Child2->SetImage(this->Image);

  this->Child1->ComputeStdDev();
  this->Child2->ComputeStdDev();
}

// Constructor sets default values
vtkImageQuantizeRGBToIndex::vtkImageQuantizeRGBToIndex()
{
  this->LookupTable = vtkLookupTable::New();
  this->NumberOfColors = 256;
  this->InputType = VTK_UNSIGNED_SHORT;

  this->InitializeExecuteTime = 0.0;
  this->BuildTreeExecuteTime = 0.0;
  this->LookupIndexExecuteTime = 0.0;
  this->SamplingRate[0] = 1;
  this->SamplingRate[1] = 1;
  this->SamplingRate[2] = 1;
  this->SortIndexByLuminance = false;
}

// Destructor deletes used resources
vtkImageQuantizeRGBToIndex::~vtkImageQuantizeRGBToIndex()
{
  if (this->LookupTable)
  {
    this->LookupTable->Delete();
  }
}

// This method is passed an input and output Data, and executes the filter
// algorithm to fill the output from the input.
// It just executes a switch statement to call the correct function for
// the Datas data types.
int vtkImageQuantizeRGBToIndex::RequestData(vtkInformation* vtkNotUsed(request),
  vtkInformationVector** inputVector, vtkInformationVector* outputVector)
{
  vtkInformation* inInfo = inputVector[0]->GetInformationObject(0);
  vtkInformation* outInfo = outputVector->GetInformationObject(0);

  vtkImageData* inData = vtkImageData::SafeDownCast(inInfo->Get(vtkDataObject::DATA_OBJECT()));
  vtkImageData* outData = vtkImageData::SafeDownCast(outInfo->Get(vtkDataObject::DATA_OBJECT()));

  outData->SetExtent(outInfo->Get(vtkStreamingDemandDrivenPipeline::WHOLE_EXTENT()));
  outData->AllocateScalars(outInfo);

  int inExt[6];
  inData->GetExtent(inExt);
  if (inExt[1] < inExt[0] || inExt[3] < inExt[2] || inExt[5] < inExt[4])
  {
    return 1;
  }

  // Input must be 3 components (rgb)
  if (inData->GetNumberOfScalarComponents() != 3)
  {
    vtkErrorMacro("This filter can handles only 3 components");
    return 1;
  }

  // this filter expects that output is type unsigned short.
  if (outData->GetScalarType() != VTK_UNSIGNED_SHORT)
  {
    vtkErrorMacro(<< "Execute: out ScalarType " << outData->GetScalarType()
                  << " must be unsigned short\n");
    return 1;
  }

  this->InputType = inData->GetScalarType();

  switch (this->InputType)
  {
    vtkTemplateMacro(vtkImageQuantizeRGBToIndexExecute<VTK_TT>(this, inData, outData));
    default:
      vtkErrorMacro(<< "Execute: This ScalarType is not handled");
      return 1;
  }

  return 1;
}

// Change the output type and number of components
int vtkImageQuantizeRGBToIndex::RequestInformation(vtkInformation* vtkNotUsed(request),
  vtkInformationVector** vtkNotUsed(inputVector), vtkInformationVector* outputVector)
{
  // get the info objects
  vtkInformation* outInfo = outputVector->GetInformationObject(0);

  vtkDataObject::SetPointDataActiveScalarInfo(outInfo, VTK_UNSIGNED_SHORT, 1);
  return 1;
}

// Get ALL of the input.
int vtkImageQuantizeRGBToIndex::RequestUpdateExtent(vtkInformation* vtkNotUsed(request),
  vtkInformationVector** inputVector, vtkInformationVector* vtkNotUsed(outputVector))
{
  // get the info objects
  vtkInformation* inInfo = inputVector[0]->GetInformationObject(0);

  int inExt[6];
  inInfo->Get(vtkStreamingDemandDrivenPipeline::WHOLE_EXTENT(), inExt);
  inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_EXTENT(), inExt, 6);

  return 1;
}

void vtkImageQuantizeRGBToIndex::PrintSelf(ostream& os, vtkIndent indent)
{
  this->Superclass::PrintSelf(os, indent);

  // Input Type is internal so we don't print it
  // os << indent << "InputType: " << this->InputType << endl;

  os << indent << "Number Of Colors: " << this->NumberOfColors << endl;
  os << indent << "Lookup Table: " << endl << *this->LookupTable;
  os << indent << "Execute Time (in initialize stage): " << this->InitializeExecuteTime << endl;
  os << indent << "Execute Time (in build tree stage): " << this->BuildTreeExecuteTime << endl;
  os << indent << "Execute Time (in lookup index stage): " << this->LookupIndexExecuteTime << endl;
}