itk中的图像分割算法（五）

今天给大家带来的是分割算法中的经典：区域生长。相信看这篇文章的你绝不会对这个词陌生的，so，具体含义本文就不解释了，直接上代码。

itk算法库中提供的类名叫：itkConnectedThresholdImageFilter

 typedef itk::ConnectedThresholdImageFilter ConnectedFilterType;
 ConnectedFilterType::Pointer connectedThreshold = ConnectedFilterType::New ();
 ConnectedFilterType->SetInput( input_image);
 ImageType::IndexType  seed0;
 seed0[0] = 75;
 seed0[1] = 142;
 //seed0[2] = 21;//这里可以设置多维的点，具体要看ImageType定义的维度
 connectedThreshold ->AddSeed(seed0);//这里可以设置多个seed的，方法就是不停的add
 connectedThreshold ->SetLower(  800 );
 connectedThreshold ->SetUpper( 3000 ); connectedThreshold ->SetReplaceValue(255);
 connectedThreshold ->Update();
output_image = connectedThreshold ->Getoutput();

是不是很简单？嗯，接口就是这样的easy。接下来看类中的具体实现。

template 
void 
ConnectedThresholdImageFilter
::GenerateData()
{
  InputImageConstPointer inputImage = this->GetInput();//输入图像
  OutputImagePointer outputImage = this->GetOutput();//输出图像

  typename InputPixelObjectType::Pointer lowerThreshold=this->GetLowerInput();//最低阈值
  typename InputPixelObjectType::Pointer upperThreshold=this->GetUpperInput();//最高阈值

  m_Lower = lowerThreshold->Get();
  m_Upper = upperThreshold->Get();

  // Zero the output  先用0填充输出图像
  OutputImageRegionType region =  outputImage->GetRequestedRegion();
  outputImage->SetBufferedRegion( region );
  outputImage->Allocate();
  outputImage->FillBuffer ( NumericTraits::Zero );
  
  //先将输入图像二值化
  typedef BinaryThresholdImageFunction FunctionType;
  typename FunctionType::Pointer function = FunctionType::New();
  function->SetInputImage ( inputImage );
  function->ThresholdBetween ( m_Lower, m_Upper );

  ProgressReporter progress(this, 0, region.GetNumberOfPixels());//这句忽略

  if (this->m_Connectivity == FaceConnectivity)//分类一：二维上的4邻域
    {
    typedef FloodFilledImageFunctionConditionalIterator IteratorType;//这里惊现FloodFill,相信一路看前面文章的同学很熟悉了
    IteratorType it ( outputImage, function, m_SeedList );
    it.GoToBegin();

    while( !it.IsAtEnd())
      {
      it.Set(m_ReplaceValue);
      ++it;
      progress.CompletedPixel();  // potential exception thrown here
      }
    }  

  else if (this->m_Connectivity == FullConnectivity)//分类二：二维上的8邻域
    {
    // use the fully connected iterator here. The fully connected iterator 
    // below is a superset of the above. However, it is reported to be 20%
    // slower. Hence we use this "if" block to use the old iterator when
    // we don't need full connectivity.

    typedef ShapedFloodFilledImageFunctionConditionalIterator IteratorType;//一个加强版本的FloodFilled
    IteratorType it ( outputImage, function, m_SeedList );
    it.FullyConnectedOn();
    it.GoToBegin();

    while( !it.IsAtEnd())
      {
      it.Set(m_ReplaceValue);
      ++it;
      progress.CompletedPixel();  // potential exception thrown here
      }
    }
#endif

}

代码不要太简单，思路就是先二值化，然后通过seed来定位，再然后floodfilled填充，然后没了~

所以，关键点在floodfilled，如果看到这里不太明白，请细翻翻我之前的文章，谢谢。

珍惜你现在拥有的一切，用真诚的笑容来面对他们，用全部的力量来保护他们，真希望这一刻就是永恒！

参考文献：

1.https://itk.org/Wiki/ITK/Examples/Broken/Images/ConnectedThresholdImageFilter