[Insight-users] 32 bit (float) TIFFs ? HDF5 ?

Matt McCormick matt.mccormick at kitware.com
Sun Jan 15 18:23:32 EST 2012


Hi Robert,

On Sun, Jan 15, 2012 at 4:24 PM,  <Robert.Atwood at diamond.ac.uk> wrote:
> Dear All:
>
> I find that attempting to input 32-bit (floating point valued) TIFF images gives me a message indicating that the ITK import method doesn't handle this. Why is this, is it especially difficult or has there just not been demand?
>

There have been recent advances in the TIFF support, but there is
always more to do.  Make sure you are using ITKv4 for the best
results.

> Though, we are migrating to HDF5 via NeXus  http://www.nexusformat.org/   hoping to get away from the inherent problem of TIFFS (too many files for the poor little data storage system, much too small limits for stacked-tiffs...) but that may have its own problems , I personally can't get the C++ library to say 'hello world' even -- what stage is HDF5 in ITK at? Here, NeXuS is the preferred flavour of HDF5, and so far the performance benefits seem to be worth the convenience costs  (few applications read it) -- so if we really go that route I'd like to try to help ensure that ITK -- therefor hopefully all applications that make use of the library ...  can read this stuff. But I'm not the programmer , I can try to recommend that our programmers lend assistance if it helps increase usability of the data form that works best for the data acquisition side; so I'd like to clarify the current hdf5 situation.

Kent Williams did a nice job of putting together an HDF5-based
reader/writer for ITK images in ITKv4.  See Modules/IO/HDF5.  You
could do something similar for the NeXuS format by translating the
NeXuS fields to their ITK counterparts.  The hard part of the HDF5
backend support is already there.  The way of doing this would be to
create a new ITKv4 Module, then submit it to the Insight Journal.

Thanks,
Matt


More information about the Insight-users mailing list