[Insight-users] Problem smoothing image 3D MRI-T1
withDiscreteGaussienImageFilter
Miller, James V (Research)
millerjv at crd.ge.com
Thu May 5 11:29:46 EDT 2005
Xavier,
The only parameters you need to set usually are the SetVariance() and
UseImageSpacingOn()/Off().
If your data an anisotropic spacing, then you should set UseImageSpacingOn().
When UseImageSpacing is on, the Variance is specified in physical units
(for instance mm). When it is off, the Variance is specified in pixels.
In the released versions of ITK, the DiscreteGaussianImageFilter
needs to be provided an image with floating point pixels. Last week,
I checked in a version to the CVS repository that will allow the
input and output pixel types to be integral.
This is likely your error. You probably gave the DiscreteGaussian
an image of ints or shorts, etc. If you are using an official release,
you need to give the filter an image of floats or doubles.
The method SetMaximumError() allows you to control the accuracy of the
filtering. The method SetMaximumKernelWidth() allows to control the
speed of the filtering (smaller kernels are faster). But you only
need to adjust these parameters when you start to address speed and
accuracy of the filtering. The defaults are all reasonable values.
Jim
-----Original Message-----
From: insight-users-bounces+millerjv=crd.ge.com at itk.org
[mailto:insight-users-bounces+millerjv=crd.ge.com at itk.org]On Behalf Of
Xavier Kite
Sent: Thursday, May 05, 2005 10:59 AM
To: insight-users ITK
Subject: [Insight-users] Problem smoothing image 3D MRI-T1
withDiscreteGaussienImageFilter
Hi,
I've already posted a message a few days ago but without answer.
I can't use this filter, I can't find the right parameters for error, var
iance and kernelwidth.
My problem is that the result is an image without
levels and with a null range of intensity.
Someone has already used t
his filter in the case of 3D images and could help me to tune it?
Than
ks for the help..
Xavier
Protek-on: CaraMail met en oeuvre un nou
veau Concept de Sécurité Globale - www.caramail.com
More information about the Insight-users
mailing list