> > Sorting transparent polygons is a very hard problem,
> > especially for complex and unpredictable shapes like
> > iso-surfaces. And note the sort order must change as
> > the user interactively rotates in 3-D. In Vis5D we
> > got around this problem by using screen-door
> > transparency, but the last time (a long time ago) I
> > tried it in Java3D it didn't work right.
> Maybe it's time to investigate this again. Any pointers on
> how to do this would be appreciated.
There is a screen door option in the Java3D API,
and the thing to do is find the corresponding call
in VisAD and experiment with changing it to screen
door. If it works, an option could be added to
GraphicsModeControl, and possibly also a new
DisplayRealType defined for ConstantMap only, as
we have done with some other GraphicsModeControl
> > I think transparency is over-rated. 3-D images are
> > ambiguous enough and using transparency just adds to
> > that ambiguity.
> There are times when in 2D images transparency is
> important. For example, if I want to overlay a radar
> image on a satellite image, I want to have portions of
> the radar image (missing data) be transparent. I can do
> this, but only if the DataReferences are adding in the
> correct order. There are many other cases where it would
> be useful to use transparency (fading images in and out like
> Tom does in many of his applets) so while I agree that for
> things like 3D displays transparency can be ambiguous, in
> the 2D displays, it is a very useful feature and currently
> a serious problem in VisAD.
Absolutely transparency has its uses, as you describe and
also in volume rendering (lots of 3-D ambiguity there, as I
wrote in my Confessions of a Visualization Skeptic column).
But its still over-rated.