Text archives Help
- From: tom fogal <tfogal@alumni.unh.edu>
- To: tlamar@bidmc.harvard.edu
- Cc: iv3d-users@sci.utah.edu
- Subject: [IV3D-USERS] Re: Re: issues with segmented data
- Date: Tue, 21 Jul 2009 11:29:18 -0600
tlamar@bidmc.harvard.edu writes:
>
Our dataset is actually not quite a series of 'flags';, rather, it
>
is continuous within our ROI and 0 outside the ROI. While nearest
>
neighbor interpolation would work, I think the ideal solution for
>
us would be if IV3d could use trilinear interpolation excluding any
>
value that was 0, or just use nearest neighbor for the perimeter of
>
the ROI and trilinear for the interior.
Ahhh, yeah, I see what you mean.
Unfortunately that's a bit harder; we could probably do this by using a
nearest neighbor texture and manually doing the interpolation ourselves
in the shader. It'd have a heavy performance hit.
The shader code is getting absurd these days; I'm not willing to add in
an option for this until we get a shader factory internally. That's
planned for 2.0 currently.
What is the shape of your ROI? If it's cubic, you could just crop your
dataset to remove all the 0-values.
>
Any idea when nearest neighbor interpolation would be added?
Well, I could say that our normal bin for features that are planned but
not yet scheduled out is "2.0". However, changing the interpolation
mode is actually pretty simple. I won't have time this week, but I'd
be very surprised if it wasn't in there a month from now.
-tom
>
-----Original Message-----
>
From: tom fogal [mailto:tfogal@alumni.unh.edu]
>
Sent: Thursday, July 16, 2009 1:51 PM
>
To: iv3d-users@sci.utah.edu
>
Subject: [IV3D-USERS] Re: issues with segmented data
>
>
Hi Tom, sorry you're having issues.
>
>
tlamar@bidmc.harvard.edu writes:
>
>
> IV3d loads the data fine, but I can't get the coloring I want. It
>
> seems to me that in computing the value for each voxel, IV3d is
>
> averaging actual values from the ROI with the zeros from outside.
>
>
Yep. ImageVis3D is doing trilinear interpolation to obtain the
>
appropriate data value within the volume. For many datasets this is
>
what one wants, but when a dataset consists of a series of `flags'
>
(i.e., "value 806 is bone, value 908 is skin", etc.) this is undesirable
>
because it `mixes' components of the datasets which should remain
>
separate.
>
>
> When I apply a 1d transfer function, the outside always takes on
>
> whatever color I try to apply to regions with low values.
>
>
>
> Do you have any suggestions?
>
>
This fix is for ImageVis3D to use nearest neighbor interpolation. This
>
is on our `todo' list, but we haven't had time for it yet.
>
>
If you're willing to dive into the code, I can overview what needs to be
>
done. Otherwise, please watch this list for release announcements which
>
mention nearest neighbor interpolation as a `new feature'.
>
>
Cheers,
>
>
-tom
Archive powered by MHonArc 2.6.16.