[Insight-users] surface textures?

[Luis Ibanez]

Hi Markus,

ITK is N-Dimensional. You can perform registration with
5D images, as long as you have enough computer memory,
CPU power and coffee suply.

Volumetric registration can certainly be done in ITK.

---

Is the 3D point-set filling a volume in space,
or is it describing a 2D surface embedded in a
3D space ?

Your problem sound pretty similar to what is done
in ITK for Finite Element Methods deformable
registration. A small region around each node in
the mesh is used to find a local translation to
the other image. The result of the registration
in this case is a displacement field.

You will find a description of deformable registration
methods in the SoftwareGuide:

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

Section 7.12, pdf-pages 230 to 236.
You may also find interesting the description of
model to image registration in presented in
section 7.14.

---

It is still not clear to me if your are managing a
3D problem, or a 2D problem embedded in a 3D space.

When we say a 3D point set, we associate 3D to the
dimension of the points, but still the cells connecting
those points can be 1D, 2D, or 3D.
In the first case you have a curve in 3D, in the second
case you have a surface in 3D, in the last case you
have a solid volume.

What is the maximum dimension of the cells connecting
the 3D points in your point-set.


Please give us clearer details of your intended
application.



Thanks


Luis



--------------------------------------

Markus Louw wrote:
> Hi Luis,
>...

> The data is a 3D pointset with a texture, (vrt file format with jpeg
> textures), and the problem is in fact a point to point correspondence
> finding problem.  Now that I think about it, my original email was rather
> vague. Sorry!
> 
> ITK at the moment canīt do volumetric registration, right?
> 
> Regardz
> 
> Markus
> 
> 
> ----- Original Message -----
> From: "Luis Ibanez" <luis.ibanez@kitware.com>
> To: "Markus Louw" <mlouw@unizar.es>
> Cc: <insight-users@public.kitware.com>
> Sent: Thursday, March 13, 2003 3:01 PM
> Subject: Re: [Insight-users] surface textures?
> 
> 
> 
>>Hi Markus,
>>
>>
>>The notion of textures in image processing is quite
>>different from its counterpart in computer graphics.
>>
>>If you mention texture in ITK context we will probably
>>think of :
>>
>>1) Simply an Image (of any dimension..., 2D, 3D, 4D)
>>
>>or
>>
>>2) The statistical characterization of gray levels
>>    in an image.
>>
>>
>>ITK has an open architecture for registraion.
>>This means that you can easily come up with new
>>registration applications by reusing basic components
>>and customizing others.
>>
>>It is likely that you could perform registration for
>>your application using ITK components, but this will
>>demand some coding effort on your part combined with
>>a moderate amount of coffee.
>>
>>---
>>
>>Could you please give us more details on the goal
>>of your application. For example:
>>
>>1) How are you representing these surface textures ?
>>
>>2) Are they 2D images remapped on the geometry of the
>>    surface ? What kind of mapping are you using ?
>>    classical mesh to mesh correspondance ?
>>
>>3) Do you want to do Rigid registration ?
>>    Are you considering changing scales ?
>>    shearing ?
>>    or full deformable registration ?
>>
>>4) Are the surfaces holding the texture having the
>>    same geometry ?
>>
>>    That is, do you have a sphere with texture to be
>>    register with an ellipsoid with texture ?
>>    or two spheres, each one with textures ?
>>
>>5) Do you want just to slide the textures in their
>>    respective geometry  ?
>>
>>    or do you want to deform their geometry altogether.
>>
>>6) How do you want to manage the dimensionality
>>    problem ?
>>
>>    That is, the textures are each one in a 2D manifold.
>>    There are very slim chances that both manifolds will
>>    be parallel anywhere in space (unless you totally
>>    share the underling geometry). In fact, the traversability
>>    theorem in singularity theory will suggest that the
>>    probability of the two manifolds to be parallel tends to
>>    zero.
>>
>>    So in priciple, the only region of 3D space where your
>>    two textures meet is a 1D curve in 3D space. Little
>>    information can be pull out of such an intersection.
>>
>>
>>
>>
>>Please let us know more details of your application.
>>
>>
>>It will be a pity if you miss the oportunity of having
>>fun with ITK.
>>
>>
>>
>>Thanks
>>
>>
>>
>>    Luis
>>
>>
>>
>>---------------------
>>
>>Markus Louw wrote:
>>
>>>Hi all
>>>
>>>My problem is to register surface textures of 3D objects.   Is ITK 1.2
>>>useful for this?  There doesnīt seem to be functionality for textures
>>>which is a pity...
>>>
>>>Markus
>>
>>
>>
>>
> 
>