More Voxels
After I stumbled upon the Voxlap Engine more or less accidentally, I couldn’t deny myself further investigation on this topic.
There has been some development by Sven Forstmann. He is working on his CUDA voxel rendering engine which is based on Ken Silvermans Voxlap algorithm.
Seems like this topic is getting more and more attention especially when we talk about sparse voxel octrees (SVO). In simple terms in this technique geometric data (voxels) is stored in octrees. The rendering process is done by tracing rays for each pixel through the octree and setting the pixel to the appropiate color of the voxel the ray intersects first. So it’s similar to traditional ray-tracing with primitives (like spheres or triangles) in continuous space. The difference is that ray-tracing works faster in discrete space (with voxels). And the main benefit is due to the volumetric data organization in octrees, which theoreticaly allow an unlimited geometrical detail and which grants a level of detail practically for free (in terms of effort).
I don’t want to go too much in detail here, but french PhD Student Cyril Crassin is one of the pioneers in this area. Also, John Carmack from id Software gave some statements on this. And there is another impressive demo by Jon Olick on youtube.
In: programming · Tagged with: CUDA, graphics programming, octree, ray-tracing, voxels, voxlap
on 2009-07-02 at 9.16 am
Permalink
“Ich möchte hier nicht zu sehr ins Detail gehen?” Zu spät ist schon passiert
on 2009-07-02 at 9.33 am
Permalink
Naja, es gibt etliche Whitepapers und Artikel über dieses Thema. Das was ich da so von mir gegeben habe ist sozusagen die Minimalessenz
Da kommen Fragen auf wie: “Wie transformiere ich eine SVO-Geometrie z.B. um einen Charakter zu animieren?” und dann wirds erst richtig spannend.
on 2009-07-05 at 10.39 am
Permalink
Spannend, Du sagst es