Image with sparse and continuous coordinates in ITK

Question

I have a raw data image which is potentially sparse and has continuous coordinates (eg 1000 pixels which are positioned on a spiral, the coordinates are floats). What is the best way to load this data into ITK for further processing and the ability to save the image in physical coordinates?

My research so far: There is itk::SpecialCoordinatesImage which I could inherit to override TransformPhysicalPointToContinuousIndex(…) and TransformPhysicalPointToIndex(…). I do not know the position and pixel number before reading the hole data stream. So for a minimal amount of speed I will need to resort the data "manually". Isn't there a better way?

Answer 1

I am more familiar with vtk than itk, so propably what comes into my mind is a bit biased. You could:

1. load the raw data into a vtk unstructured grid (see for example the function ReadFinancialData in http://vtk.org/gitweb?p=VTK.git;a=blob;f=Examples/Modelling/Cxx/finance.cxx )
2. then voxelize it to an image. For example. see http://www.vtkjournal.org/browse/publication/713 (I've never used it, I dont' know if it is compatible with the last versions) or http://www.vtk.org/Wiki/VTK/Examples/Cxx/PolyData/PolyDataContourToImageData