[Insight-users] (no subject) : ITK and Microscopy Stitching / Mosaics

[James Carroll]

Thank you Luis,

We actually already have the registration, tiling and mosaics working
beautifully, but I would love to leverage some of the progressive (I
call it pull-model, even though I'm pushing out to a file in this case)
ROI-based image handling.

The image sizes are also not a problem, check out:

Dolphin brain:    http://tinyurl.com/ytuf9

The dolphin brain would be 77GB of image information if it was
uncompressed. Zoom in with the (+)=3D magnifying glass about ten times =
to
see individual cells.  It takes up just about 4GB of hard drive space,
and is usable because we only serve one field of view at a time.  We do
image processing in the Java viewer on the client side to avoid
processing the whole image.

What I want to do is write the information using several methods in
addition to what we do now.  I'm excited to try IORegion, and
itk::TIFFImageIO::Write.  A colleague at RPI has been encouraging us to
use the ITK for some time now and this is a good excuse.  I'm hoping to
use imageIOBase as the basis for additional implementations of various
huge image file formats in the coming months.  Not just for the web
microscopy, but for several quantitative projects.=20

I will look into the Meta Image format, although we do need to store
things in single files.  Single files are more convenient to handle, and
they avoid some patent landmines in the microscopy world.  (Although I
do understand TIFF limits us to files smaller than 2GB compressed.)
GZip would be nice too, but I think PNG-in-TIFF does something similar
in spirit.

-Jim




Hi James,

You are not alone !   :-)

Almost everybody doing Microscopy these days is
having the same problem.

It is feasible to do stitching or mosaics in ITK.
It will require however some effort on your part
since the original design of the image registration
framework was mostly oriented to images with large
overlapping regions.

The first thing you will want to experiment with
is the masking functionality of the image metrics.
For example in the MeanSquareImageMetric.

If you are not yet familiar with the Registration
framework in ITK, you may benefit from reading
the chapter on ImageRegistration and the chapter
on ImageResampling from the SoftwareGuide

   http://www.itk.org/ItkSoftwareGuide.pdf

There is probably a critical mass of ITK users
dealing with Microscopy for justifying the
creation of a number of classes supporting
2D Stitching and Mosaics. Note that users doing
satellite imagery have exactly the same need.

BTW, I don't think you really want to physically
compose the mosaic at high resolution. Having an
image of 1.5Gb will not be very useful since you
will face a good number of difficulties for
visualizing it and processing it.  You are probably
better off performing the image registration between
neighborhood images and storing the transforms along
with the images.  Then creating a visualization application
(VTK will be ideal for this) where you load only the
images that you are currently viewing an apply the
transforms to them, plus some blending (image fusion)
visualization techniques.

Instead of TIFF I would suggest you to use the MetaImage
format which now supports gzip compression and allows
you to store the spatial transform with the images,
in this way you don't have to resample the images for
storage, and don't have to compose the actual mosaic
in a huge file.



Please let us know if you have further questions.


   Thanks



     Luis


----------------------
James Carroll wrote:

> Hi,
>=20
> I'm about to spend some time using itk::TIFFImageIO to write large
tiled
> tiffs,  setting the region of interest, then feeding it lots of data.
> I'm hoping I can incrementally write tiffs that are jpeg-in-tiles
> compressed, and have a total size over 1.5 GB. =20
>=20
> Is this a reasonable thing for me to try in ITK?  Anyone have any
> similar examples?  (C++ or Python both welcome)
>=20
> Thanks,
> -Jim
>=20
>=20
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>=20



_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users