[vtkusers] The vtkMINCImageReader - second iteration

[David Gobbi]

Hi All,

I've committed some significant enhancements for VTK minc support.  The 
biggest enhancement is a vtkMINCImageWriter for writing minc files from VTK.

There is also a class called "vtkMINCImageAttributes" that stores all 
the attributes of the MINC file.  By using this class, you can easily 
pass all of the attributes from the reader to the writer so that patient 
information, dimension order, and other important information is 
preserved.  The writer has a SetHistoryAddition() method for adding to 
the history attribute of the minc file.

Here is a short summary of what these classes do, and what they don't 
do.  First the good stuff:

1) The reader automatically reorders the data into VTK x,y,z order.
2) The reader optionally converts real-valued integer data to floating 
point (use RescaleRealValuesOn() )
3) The reader can handle 4D data: use SetTimeStep() to choose which time 
frame to read.
4) The reader supports data caching, i.e. it only reads and stores the 
specified UpdateExtent into memory.
5) All minc attributes can accessed via the vtkMINCImageAttributes class
6) The writer can write data in arbitrary dimension order (set dimension 
order in vtkMINCImageAttributes)
7) The writer will update the pipeline slice-by-slice, making it 
possible to process data without loading the entire data volume into 
memory at once.
8) The writer can validate the file attributes before they are written 
to ensure that they meet the minc standard

And the bad:

1) Only minc 1.0 files are supported, not hdf5-based minc 2.0 files
2) Irregular dimension spacing is not supported.   I might add support 
for irregular time spacing.
3) Dimension widths are not supported (does anyone actually use these?)
4) The reader does not do ICV-like data type conversion or dimension 
reordering.

This is the last big chunk of work that I will be doing on these classes 
for a while.  The only future enhancement that I have planned is 
incorporating VTK's new time handling into the reader and writer.  This 
would allow  VTK pipelines to automatically carry 4D data from the 
reader to the writer by processing each time step of the file in turn.  
It would also, I suppose, allow MINC files to be used as 4D data sets in 
ParaView.

Any comments are welcome.

 - David