Re: Performer Rendering Problem?

New Message Reply Date view Thread view Subject view Author view

Scott Herod (herod++at++aspen.evt.com)
Tue, 31 Aug 1999 12:18:43 -0600


This sounds like a problem similar to what I call "Alpha-Cutting".
You didn't mention whether your two geosets are declared to be
transparent. If they are not, then this probably isn't your problem.
Your comments about the two rectangles being close to each other
and only relatively the same size is what makes me think of this.

Basically, Alpha-Cutting occurs when you have two transparent
GeoSets very close to each other. They don't have to be co-planar
but it is most noticable when they are. Furthermore, the center
of the front GeoSet must actually be further from the camera than
that of the back GeoSet. ( That can happen if they are translated
sidewards with respect to each other. ) Since Performer sorts
transparent GeoSets by distance from the camera before drawing,
the front GeoSet is drawn first ( since it is further from the
camera based on its center ). However, since the front GeoSet
is in front of the back one, the z-buffer is already filled when
it comes time to draw the back GeoSet so it is not completely filled
and any transparency in the front GeoSet is blended with the
background rather than with the back GeoSet. The only way I know
to fix this problem is to write your own back to front sorting
algorithm that is specialized to your particular data set.

I can tell you more about how I've fixed special case of the above
if you wish.

Scott Herod
scott++at++evt.com

Neil Witcomb wrote:
>
> ** High Priority **
>
> I have created two pfGeode which have a single plane as their geoset. They are co-planar and are placed 0.1 meters apart. They are both attached to the same pfDCSand are relatively the same size. When I move the camera close to the planes, looking head on, the smaller one doesn't disappear but the ground/sky is rendered in it's position, even if something is behind it. When the camera is moved up and down the horizon moves as expect.
>
> Does anyone know what is happening?
>
> Thanks
> Neil Witcomb
>


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Tue Aug 31 1999 - 11:18:57 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.