[Insight-users] Re: VersorRigid3DTransform
Radhika Sivaramakrishna
Radhika Sivaramakrishna" <radshashi at earthlink.net
Wed, 31 Dec 2003 12:36:33 -0800
Hi Luis,
Thanks for the explanation. It was helpful.
I had some further questions.
How do I set the center of rotation to something other than the origin? I
can do this by explicitly setting the values in finalParameters but how do I
do it using versors and setting the rotation angle explicitly?
Also is the optimization done in a 9-parameter space including center of
rotation or is it a 6-parameter space?
Also, when the offset is printed out (for example in the example given under
CenteredTransform), is this always given as though the rotation were about
the origin rather than the center of rotation?
Is there a way to print out the true translation?
Radhika
----- Original Message -----
From: "Luis Ibanez" <luis.ibanez at kitware.com>
To: "Radhika Sivaramakrishna" <radshashi at earthlink.net>
Cc: <insight-users at itk.org>
Sent: Wednesday, December 31, 2003 11:42 AM
Subject: Re: VersorRigid3DTransform
>
> Hi Radhika,
>
> A) The memory problem that you are encountering is
> due to the fact that the "ParameterType" used by
> ITK transform for representing the array of
> parameters is indeed a vnl_vector<>.
>
> This is a variable size vector that by default has
> size=0. You must replace the declaration:
>
> TransformType::ParametersType finalParameters;
>
> with something like:
>
> TransformType::ParametersType finalParameters(
> transform->GetNumberOfParameters() );
>
> or something like:
>
> TransformType::ParametersType finalParameters =
> transform->GetParameters();
>
> In both cases the vector-size of finalParameters
> will take the correct value.
>
>
>
> B) The value that you provided for the parameter[0]
> doesn't correspond to 10 degrees. A Versors is
> a unit quaternion. The magnitude of the three
> versor components is equal to sin(angle/2).
>
> You probably computed
>
> parameter[0] = 0.174532 = sin( 10 )
>
> while you should do
>
> parameter[0] = 0.087155 = sin( 10 / 2 )
>
>
> C) Yes there is a better way of initializing the transform.
> Please look at the methods of the Versor class:
> http://www.itk.org/Insight/Doxygen/html/classitk_1_1Versor.html
>
> You can simply do:
>
> TransformType::VersorType versor;
> versor.SetRotationAroundX( 10.0 * PI / 180.0 );
> transform->SetRotation( versor );
>
> or you could do
>
> TransformType::AxisType axis;
> axis[0] = 1.0;
> axis[1] = 0.0;
> axis[2] = 0.0;
>
> transform->SetRotation( axis, 10.0 * PI / 180.0 );
>
>
>
> Please let us know if you have further questions,
>
>
> Thanks
>
>
> Luis
>
>
> -------------------------------
> Radhika Sivaramakrishna wrote:
>
> > Hi Luis,
> > I was trying to use the VersorRigid3DTransform to artificially transform
a
> > given image by 10 degrees about the center of the image only in the
> > X-direction (ie no rotation in Y and Z) and no translation either. I am
> > doing this to get a better understanding of this transform because I
will
> > need to use it to register a large number of unknown cases.
Unfortunately, I
> > am running into some problems.
> > I modified the example in ImageRegistration8.cxx to do this.
> >
> > Here is what I did:
> >
> > I read in the fixed and moving image (in my case both are from the same
> > file). Since no registration is required this time, I only defined a
> > Transform of type VersorRigid3DTransform.
> >
> > I then set the finalparameters of my transform directly in the following
> > way:
> >
> > TransformType::ParametersType finalParameters;
> >
> >
> >
> > finalParameters[0] = 0.174532;
> > finalParameters[1] = 0;
> > finalParameters[2] = 0;
> > finalParameters[3] = 128.0;
> > finalParameters[4] = 128.0;
> > finalParameters[5] = 82;
> > finalParameters[6] = 0;
> > finalParameters[7] = 0;
> > finalParameters[8] = 0;
> >
> > since I want a 10 degree rotation in X, my image is of size 256x256x164.
> >
> > I then did the usual resampling and casting to create my final image.
> > However a memory error is reported when I try to set the finalParameters
> > directly. Can you tell me what the problem is?
> >
> > Is there a better way of artificially creating a known transformation?
> >
> > Also, when I get to the actual image registration stage, what is the
best
> > optimizer for this transform? Is it the VersorTransformOptimizer?
> >
> > Thanks
> > Radhika
> >
> >
> >
>
>
>