3D viewer & mesh averaging

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

3D viewer & mesh averaging

Gabriel Landini
Hi,
After loading 2 or more meshes (displayed as surfaces) into the 3D viewer and
registering them, is there a way of finding their average? (ie create a new
mesh that is the result of the blending/average of the others).

If not, does anybody know of a program that can do this?
Cheers

Gabriel
Reply | Threaded
Open this post in threaded view
|

Re: 3D viewer & mesh averaging

Albert Cardona-2
2011/7/21 Gabriel Landini <[hidden email]>:
> Hi,
> After loading 2 or more meshes (displayed as surfaces) into the 3D viewer and
> registering them, is there a way of finding their average? (ie create a new
> mesh that is the result of the blending/average of the others).
>
> If not, does anybody know of a program that can do this?


Tough one. I would find uses for such a method if it exists.

Albert



--
http://albert.rierol.net
http://www.ini.uzh.ch/~acardona/
Reply | Threaded
Open this post in threaded view
|

Re: 3D viewer & mesh averaging

dscho
Hi Albert & Gabriel,

On Thu, 21 Jul 2011, Albert Cardona wrote:

> 2011/7/21 Gabriel Landini <[hidden email]>:
>
> > After loading 2 or more meshes (displayed as surfaces) into the 3D
> > viewer and registering them, is there a way of finding their average?
> > (ie create a new mesh that is the result of the blending/average of
> > the others).
> >
> > If not, does anybody know of a program that can do this?
>
> Tough one. I would find uses for such a method if it exists.

The real problem is that the "average" is not well defined in that case...
if you think about two images with two circles, one time side by side and
the other one on top of the other, what is the average? Is it on circle in
the center? Is it two circles in a diagonal arrangement (and which
diagonal)?

Having said that, one strategy that might work is how we interpolate
segmentations in Fiji's Segmentation Editor: first make binary images from
the segmentations (could be done from meshes, too), then apply the
Euclidean Distance Transform to both binary images, and then interpolate
the transformed images. After that, transform back and reconstruct
segmentations (or in your case, meshes).

In the scenario I gave above, this would most likely result in an
"average" that is one bigger, central circle. Not what I would expect to
be what you want :-)

Ciao,
Dscho