Re: VisAD and transparency

• To: JR Schmidt <schmidt@xxxxxxxxxxxxx>
• Subject: Re: VisAD and transparency
• From: Bill Hibbard <billh@xxxxxxxxxxxxx>
• Date: Thu, 21 Apr 2005 15:22:53 -0500 (CDT)

Please always send technical questions to the list rather
than just to me - I'm not wlays the best source of technical
information.

If you look at the detail of TRANSPARENCY_SORT_GEOMETRY,
its as I described: it sorts Geometries relative to each
other, but does not sort triangles within Geometries. For
this to work with iso-surfaces, you'd need almost as many
Geometries as trianlges. A BIG efficiency hit.

Bill

On Thu, 21 Apr 2005, JR Schmidt wrote:

> > Iso-surface triangles (and triangles from other sources)
> > in VisAD get sewn together in long polytrianlge strips
> > which greatly increase efficiency, and these strips are
> > in general not compatable with sorting. As I recall, Java3D
> > does provide a way to segregate triangles into multiple
> > Geometries that it will then sort in depth order, but for
> > this to work with complex shapes like iso-surfaces you'd
> > have to segregate into a very large number of Geometries.
> >
> > It is possible that you may find a way to do what you want,
> > but it will be a lot of work and probably result in slower
> > rendering and greater memory use.
>
> It seems as if Java3D does implement the required sorting of transparent
> items in version 1.3.  For example, see:
> http://java.sun.com/products/java-media/3D/forDevelopers/J3D_1_3_API/j3dapi/
> and look at the 'View' class.
>
> As you can see from that page, Java 3D v1.3 added the
> 'setTransparencySortingPolicy' method to this class, which seems to do
> exactly what would be required.  I tried setting this to
> 'TRANSPARENCY_SORT_GEOMETRY' from with my VisAD app:
>
>             DisplayRenderer dRenderer = display.getDisplayRenderer();
>             View view = ((DisplayRendererJ3D) dRenderer).getView();
>             view.setTransparencySortingPolicy(
> View.TRANSPARENCY_SORT_GEOMETRY );
>
> but it didn't seem to help.  Perhaps the problem, as you mentioned, is that
> VisAD has created long polytriangle strips.  This would seem consistent with
> the 'banding' I am seeing in my image.  Perhaps a each 'band' in my image is
> in fact a strip of polygons.
>
>     JR
>
>