Am Fri, 27 Apr 2007 16:54:07 -0400 schrieb Warp:

> Lukas Winter <web### [at] removeitgeloeschtnet> wrote:
>> POV has no real support for any
>> LOD-techniques (level of detail). That is not so important for
>> raytracers,
> 
>   What makes you think so? LOD filtering is equally important in
>   raytracing
> as it is with any other rendering technique. The problem is that it's more
> difficult to implement in raytracing because of all the reflections and
> refractions and such.
> 
>   The movie Cars was raytraced (instead of scanline-rendered as usual),
> and they used mip-mapping and LOD filtering. They used some type of ray
> volume algorithms to determine the mipmap levels. It's not trivial, but
> possible.

With a scanline renderer, the times decrease more trastically when you
lower the polygon count. You can also circumvent LOD-techniques for
textures with a raytraces by doing a lot of supersampling (like depth of
field). If I implemented a raytraces (which I partly dream of and partly
already have done) I would try to include such things (by keeping a cone 
along with each ray or each bundle of rays). But it's not a must, as there
are other features in raytracers that make up for it.

Post a reply to this message