[Insight-users] Evolutinary Optimizer

Lydia Ng lydiang at gmail.com
Sun Jun 18 01:15:06 EDT 2006


Hi Jonathan,

One thing you should keep in mind when modifying the code is that the
transform parameters you are optimizing are the four components of a
quaternion and three translation components.

A quaternion is only purely rigid if the magnitude is one. One of the
feature of the QuaternionRigidTransformGradientDescentOptimizer is that this
normalization is done at each step.

If you swap out the optimizer and unless you do something special (e.g. you
have set up and observer to the optimizer and is doing the normalization
yourself), the magnitude of the quaternion might drift and you will no
longer have a rigid transform.

To ensure you end up with a purely rigid transform you will either have to
do the normalization at each step or use another type of rigid transform
such as Euler3DTransform.

- Lydia



On 6/14/06, Jonathan Wong <jon.the.wong at gmail.com> wrote:
>
> Hi
>
> I currently modified my the MultiResMIRRegistration example and have
> gotten it to compile by simply changing the optimizer from Quaternion to the
> OnePlusOneEvolutionary optimizer.
>
> I used the default settings everyone seems to be using.
>
>   m_Optimizer->Initialize( 10,1,-1 );
>   m_Optimizer->SetEpsilon( 1.0 );
>   m_Optimizer->SetMaximumIteration( 4000 );
>
>
> and i checked my scales, where the translation scale is 16mm, since my
> image is 263 with a pixel size of 0.062 mm.
>
> I keep on getting the famous "all the sampled point mapped to the outside
> o the moving image" error.
>
> I've been reading similar posts on the mailing list, and they always talk
> about setting the radius, shrink, and grow parameters as well as scaling.
>
> However, I wanted to know what Epislon is for? Thanks.
>
>
> Jonathan Wong
>
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20060617/9a3867d3/attachment.html


More information about the Insight-users mailing list