Anisotropic Diffusion Filter - Source Code

michael · February 17, 2020, 8:42am

Hi there! New to ITK, so it’s possible I’m missing something obvious, but I have been searching through the documentation for the GradientAnisotropicDiffusionFilter in Simple ITK and have found the descriptions of the anisotropic diffusion function (https://itk.org/Doxygen/html/classitk_1_1AnisotropicDiffusionFunction.html#abdd1b2f3a6e99c69c6f5582f3d266062) and some code for the filter itself (https://itk.org/SimpleITKDoxygen/html/classitk_1_1simple_1_1GradientAnisotropicDiffusionImageFilter.html), but I can’t seem to find the code where I can review the math being done. Could someone point me to it?

dzenanz · February 17, 2020, 3:16pm

The math should be here:

github.com

InsightSoftwareConsortium/ITK/blob/344e9bc58e69ed921f540c67d87dbb66fee4c466/Modules/Filtering/AnisotropicSmoothing/include/itkGradientNDAnisotropicDiffusionFunction.hxx#L100-L154


// Calculate the centralized derivatives for each dimension.
for (i = 0; i < ImageDimension; i++)
{
  dx[i] = (it.GetPixel(m_Center + m_Stride[i]) - it.GetPixel(m_Center - m_Stride[i])) / 2.0f;
  dx[i] *= this->m_ScaleCoefficients[i];
}


for (i = 0; i < ImageDimension; i++)
{
  // "Half" directional derivatives
  dx_forward = it.GetPixel(m_Center + m_Stride[i]) - it.GetPixel(m_Center);
  dx_forward *= this->m_ScaleCoefficients[i];
  dx_backward = it.GetPixel(m_Center) - it.GetPixel(m_Center - m_Stride[i]);
  dx_backward *= this->m_ScaleCoefficients[i];


  // Calculate the conductance terms.  Conductance varies with each
  // dimension because the gradient magnitude approximation is different
  // along each  dimension.
  accum = 0.0;
  accum_d = 0.0;

This file has been truncated. show original