[Insight-users] Question regarding BSplineScatteredDataPointSetToImageFilter

[Nicholas Tustison]

Okay, here's some pseudocode which hasn't been tested
since you have to provide it the interface to your values 
but you should get the general idea.

  typedef float RealType;
  typedef itk::Image<RealType, ImageDimension> RealImageType;

  // Read in your reference image which will define the domain of 
  // your B-spline displacement field.  Assume, for simplicity, that it
  // has identity direction

  typedef itk::ImageFileReader<RealImageType> ImageReaderType;
  typename ImageReaderType::Pointer reader = ImageReaderType::New();
  reader->SetFileName( XXXX );
  reader->Update();

  typedef itk::Vector<RealType, ImageDimension> VectorType;
  typedef itk::Image<VectorType, ImageDimension+1> TimeVaryingDeformationFieldType;

  typedef itk::PointSet<VectorType, ImageDimension+1> TimeVaryingDeformationFieldPointSetType;
  typename TimeVaryingDeformationFieldPointSetType::Pointer fieldPoints =
    TimeVaryingDeformationFieldPointSetType::New();
  fieldPoints->Initialize();
  unsigned long count = 0;

  // Assume points are stored in a 2-D matrix called 'mySamplePoints' where the column
  // is the time point and the row is the point

  for( unsigned int t = 0; t < numberOfTimePoints; t++ )
    {
    for( unsigned int n = 0;  n < numberOfPointsPerTimePoint; n++ )
      {
      VectorType displacement = mySamplePoints[n][t] - mySamplePoints[n][0];
      
      TimeVaryingDeformationFieldPointSetType::PointType parametric point;
      for( unsigned int d = 0; d < ImageDimension; d++ )
        {
        point[d] = ( mySamplePoints[n][0] )[d];
        }
      point[ImageDimension] = t;

      fieldPoints->SetPoint( count, point );
      fieldPoints->SetPointData( count, displacement );
      count++;
      }
    }   

  TimeVaryingDeformationFieldType::PointType origin;
  TimeVaryingDeformationFieldType::SpacingType spacing;
  TimeVaryingDeformationFieldType::SizeType size;

   for( unsigned int d = 0; d < ImageDimension; d++ )
      {
      origin[d] = reader->GetOutput()->GetOrigin()[d];
      size[d] = reader->GetOutput()->GetLargestPossibleRegion().GetSize()[d];
      spacing[d] = reader->GetOutput()->GetSpacing()[d];
      }
  // Now include the temporal information.  You can change this to whatever
  // resolution you like.  You just need to make sure that 
  //  ( size[ImageDimension] - 1 ) * spacing[ImageDimension] = ( numberOfTimePoints - 1)
  // 
  origin[ImageDimension] = 0;
  size[ImageDimension]  = numberOfTimePoints - 1;
  spacing[ImageDimension] = 1;

  typedef itk::BSplineScatteredDataPointSetToImageFilter
    <TimeVaryingDeformationFieldPointSetType, TimeVaryingDeformationFieldType> BSplineFilterType;
  typename BSplineFilterType::Pointer bspliner = BSplineFilterType::New();
  bspliner->SetOrigin( origin );
  bspliner->SetSpacing( spacing );
  bspliner->SetSize( size );
  bspliner->SetGenerateOutputImage( XXXX );
  bspliner->SetNumberOfLevels( XXXX );
  bspliner->SetSplineOrder( XXXX );
  bspliner->SetNumberOfControlPoints( XXXX );
  bspliner->SetInput( fieldPoints );
  bspliner->Update();




On Jan 4, 2012, at 2:52 AM, Kerstin Müller wrote:

> Hi,
> 
> yes correctly.
> 
> 2012/1/3 Nicholas Tustison <ntustison at gmail.com>
> Okay, I'm assuming that these points correspond in 
> time, correct?  For example, point 937 in time point 0
> corresponds to point 937 in time point 1, 2, 3, ...133,
> right?  
> 
> 
> 
> On Jan 3, 2012, at 11:55 AM, Kerstin Müller wrote:
> 
>> Hi,
>> 
>> the basic problem I wanna solve:
>> 
>> I'll have scattered points in 3D which vary over time, they describe a motion of a surface over time. They are not ordered in x-y-and z-dimension.
>> Now I want to represent that motion by BSplines in order to generate a dense motion vector field defined on specific 3D voxel positions.
>> I'll have 133 timesteps and ~960 sample points in each time step. Now I want to fit the BSpline to it and resample it.
>> 
>> All the best,
>> 
>> Kerstin
>> 
>> 2012/1/3 Nicholas Tustison <ntustison at gmail.com>
>> In that case, let's start with the basic problem set-up.
>> Before, you described your problem as follows:
>> 
>> 
>>>> I'm using the BSplineScatteredDataPointSetToImageFilter from the Insight Journal. However,
>>>> I cannot find the correct parameter in order to make the filter work and does not crash during the evaluation.
>>>> I'll have scattered 3-D points over time and I want to fit a 4-D Bspline field to that points. Afterwards I want to evaluate the Bspline on a dense volume grid to one time step.
>> 
>> 
>> What do these scattered 3D+t points represent?  
>> Is it a curve, a time-varying scalar field, or something
>> else?
>> 
>> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20120104/cb222684/attachment.htm>