Posted by
Hadjilucas, Lucas on
Jun 26, 2014; 7:30pm
URL: http://imagej.273.s1.nabble.com/HDF5-support-tp5008443p5008467.html
Hi Antoine,
The actual reason for using hdf5 is because I have some very large arrays I need to pass from Matlab -> Fiji.
I tried using Miji with the command CreateImage to instantiate a new image from within Matlab but because the array is so big I get Java Heap Out of Memory errors. This is despite having adjusted the Matlab Java heap space and having enough memory on my machine. It is also painfully slow to instantiate a large array with MIJ.createImage. For some reason it appears to be faster to just write the data on disk and load it with MIJ.run('Open...', 'path=[filepath]')
I used to use FITS for writing the large volume as a single file but the FITS libraries on windows Matlab have an issue reading files over 4GB (despite being 64-bit). Therefore I worked around that by replacing FITS with the HDF5 format which Matlab appears to be more comfortable with reading/writing when it comes to file size of 4GB+.
The problem now is that ImageJ cannot read in that HDF5 data...
Kind regards,
Lucas
-----Original Message-----
From: ImageJ Interest Group [mailto:
[hidden email]] On Behalf Of Antoine Bergamaschi
Sent: 26 June 2014 15:03
To:
[hidden email]
Subject: Re: HDF5 support
Hi,
The HDF group has developed a software named Hdf5 view <
http://www.hdfgroup.org/products/java/index.html>. You can use this soft to read and extract your dataset and then use imageJ to analyse the data.
++
Antoine Bergamaschi
2014-06-26 15:37 GMT+02:00 Mario Emmenlauer <
[hidden email]>:
> Hi,
>
> I think the question is how your HDF5 file is internally formatted.
> HDF5 has not many constraints an internal structure, so it leaves full
> freedom to you, to store your datasets in different internal paths or names or ...
> It can not know, for example, how multiple channels would be stored,
> if you have that, or how your x/y/z dimensions are stored, if you have that.
> All this is free to decide when writing the file, and no reader can
> know it in advance.
>
> So there is not really a "plain" HDF5 reader because it would need to
> know at least something about your internal structure of the file.
> If you have that, it should be trivial to adjust one of the existing
> readers to read your files.
>
> All the best,
>
> Mario
>
>
>
> On 26.06.2014 13:15, Hadjilucas, Lucas wrote:
> > Hi,
> >
> > I was wondering if there is a way to read in an hdf5 (.h5) dataset
> > into
> ImageJ.
> >
> > I already tried using the plugin in the page below but it seems not
> > to
> work for me for files above 4GB in size. I am using win7 64-bit along
> with 64-bit imagej&java so there should not be a 32-bit limitation in
> that respect.
> >
> >
>
http://lmb.informatik.uni-freiburg.de/resources/opensource/imagej_plug> ins/hdf5.html
> >
> > I had a quick look in the bioformats plugin but could not find
> > anything
> compatible to plain hdf5 format.
> >
> > Kind regards,
> > Lucas
> >
> > --
> > ImageJ mailing list:
http://imagej.nih.gov/ij/list.html> >
>
> --
> A: Yes.
> > Q: Are you sure?
> >> A: Because it reverses the logical flow of conversation.
> >>> Q: Why is top posting annoying in email?
>
> Mario Emmenlauer BioDataAnalysis Mobil: +49-(0)151-68108489
> Balanstrasse 43 mailto: mario.emmenlauer * unibas.ch
> D-81669 München
http://www.marioemmenlauer.de/>
> --
> ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html