Hi all, Has anyone over here tried using occlusion beams in combiantion with octrees to perform a scene graph function in a 3D engine? I have the idea that it could work fine, but I don''t know exactly where and how to start. What I mean is this: (ascii probably gets screwed) A view: < A B A So, object A would block B in this case. No need to render B. We could determine this by thinking of ''view'' as a light. All objects that are in shadow, are invisible. Now, I was thinking about marking octree nodes as being totally opaque from certain view angles, so I can derive a set of planes out of it. All objects inside the volume defined by these planes, and behind the ''occluder object'' is invisible. Is this a feasible approach? Has anyone experimented with this? Thanks in advance, DaBit

There was an article an Gamasutra about that a few months ago. I believe it was actually a chapter from a book covering various culling methods.

I think this is the book you want

http://www.realtimerendering.com

Okay, I did order the book at Amazon. It is time to stop reinventing the wheel.

DaBit.