meshing 3d point cloud in ImageJ ?

[Bill Christens-Barry]

Does anyone know of java code that could be used to create a plugin for constructing tetrahedral
volume meshes or 3d triangular surface meshes from 3d point cloud data? I'd like to do surface area
and volume calculations from such data.

The objects for which I have point cloud data have surfaces containing concave regions, so meshes
based on the convex hull are not appropriate. Also, the coordinates of the points in the point cloud
are arbitrary (within some bounding box), which means they don't easily sort into uniformly spaced
slices. In other words, if I assign each point to a slice in a stack, then I need prohibitively many slices
if I'm to retain the precision of the measured values. Typically, I have data spanning 10's of
centimeters in each dimension, and have a precision of better than 100 micrometers. Thus, a stack
spanning 10 cm x 10 cm x 10 cm would require a stack that is 1000 x 1000 x 1000, which is not a
pretty picture for subsequent processing with only 2 GB of ram.

Thanks.

Bill Christens-Barry

[Jonathan Hilmer]

I'm not familiar with any Java implementations of the algorithms that
accomplish what you are looking for.  This might be your best option,
outside of working in other programs or languages:
http://ij-plugins.sourceforge.net/ij-vtk/index.html

Jonathan

On 9/20/07, Bill Christens-Barry <[hidden email]> wrote:

> Does anyone know of java code that could be used to create a plugin for constructing tetrahedral
> volume meshes or 3d triangular surface meshes from 3d point cloud data? I'd like to do surface area
> and volume calculations from such data.
>
> The objects for which I have point cloud data have surfaces containing concave regions, so meshes
> based on the convex hull are not appropriate. Also, the coordinates of the points in the point cloud
> are arbitrary (within some bounding box), which means they don't easily sort into uniformly spaced
> slices. In other words, if I assign each point to a slice in a stack, then I need prohibitively many slices
> if I'm to retain the precision of the measured values. Typically, I have data spanning 10's of
> centimeters in each dimension, and have a precision of better than 100 micrometers. Thus, a stack
> spanning 10 cm x 10 cm x 10 cm would require a stack that is 1000 x 1000 x 1000, which is not a
> pretty picture for subsequent processing with only 2 GB of ram.
>
> Thanks.
>
> Bill Christens-Barry
>

[Marcel]

Why don't you use the OpenGl bindings to archieve this.
If you need to construct an 3d plugin this is the easiest solution.

Available at:
https://jogl.dev.java.net/
The bindings can easily be integrated in ImageJ.

With kind regards
Marcel