PDA

View Full Version : rendering unnessecary stuff



wavk
08-14-2003, 04:47 AM
Hi,

I've got a rendering problem.

I like the way how lightwave renders polygon aftre polygon, but I don't like the fact that it renders unnecessary polygons. In this project I'm working on, it first renders the entire glass panel wall, which takes a LONG time with radiosity. And after that, it renders the concrete walls and window frames over them, which leaves about 5% of the glass still visible. After that, the concrete floor is rendered, which is then covered 100% by a wooden floor. I think this render could be rendered about 10 times as fast without rendering the unnecessary parts.

Then I long back to yafray, which just renders pixel by pixel.

Does this problem have anything to do with the bounding boxes of the objects? I tried cutting out only the windows which were visible and put them in another layer, but that didn't seem to work.

Any ideas?

Wybren van Keulen
Funny Farm

colkai
08-14-2003, 04:56 AM
If your window poly is large, just sub-divide it (faceted) a couple of times so it is made up of smaller polys - it helps LW decide what is actually "at the front".

I found this can really cut into render times for certain scenes.

wavk
08-14-2003, 05:21 AM
Hi colkai,

Thanks for the response. That sounds like a logical thing, but it's not the problem in this scene, for the glass panel walls consists of hudreds of glass panels, each seperate quads. Any other ideas?

Wybren

Mylenium
08-14-2003, 09:16 AM
You cannot change that if you are using radiosity. Radiosity is global whereas ray tracing is local. What does this mean?

For radiosity rays are fired directly from light sources, luminous objects etc. depending on a variety of factors these can be very many. It than analyzes their behaviours (intersection with object = visible) and fires even more rays to determine color, transparencies, reflections, color bleeding etc.. Since radiosity algorithms are "dumb", they do this without considering which part is currently visible in the camera. Ray tracing does it exactly the other way around. Rays are fired from the camera, so you normally control the number of rays with the camera/ rendering properties (resolution being one of the most noticeable contributing factors).

Regardless of this theoretical hubub there are some reliable solutions to your problem:

a) Don't use radiosity.

b) Remove your unnecessary objects/ parts. This may however drastically affect the overall light distribution in your scene and ruin your look.

c) Replace your large object with a textured dummy. Use Texture Baker to bake a color/ luminosity/ diffuse texture for your object and then map it onto a simplified object. Since there are fewer possible intersections but certain info is stored in the texture map this will acelerate rendering.

Hope that made some sense to you.

Mylenium

wavk
08-18-2003, 03:31 AM
Hi Mylenium,

Thanks for your lengthy expenation. I don't agree, however. For it renders exactly the same using this method:

http://vbulletin.newtek.com/showthread.php?s=&threadid=9683

Wybren