"Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> "Grassblade" <nomail@nomail> wrote:
> > I used the SDL raytracer found in the help file
> > (http://www.povray.org/documentation/view/3.6.1/123/) to refine an idea
> > more or less suggested by Bruno Cabasson.
> > Top picture skips a ray if (square of) distance in colour-space is smaller
> > than an user supplied threshold.
> > Middle pic is the result of the unadulterated SDL raytracer.
> > Bottom pic is the difference of the two.
> > The supplied Skip-a-ray pic requires about 65-70% of the time to render.
>
> The fact is  that, basically and in principle, you can't predict whether or
> not an object happens to occlude another at the next pixel or next ray.
Well, actually making predictions based on past history is the job of
statistics. In general you can't have a 100% accurate predictor, though;
which I believe is what you wanted. So, yes, it is possible to get an
imperfect predictor on the first pass. But the math involved can get really
ugly.
 My
> suggestion was about the fact that in ALREADY RENDERED images, entire areas


>
> It appears now that the answer is no.
>
> But I am still wondering if a first 'geometrical' pass at render-time could
> help collect data (like direction-object correspondance) that could be used
> FOR SUBSEQUENT RENDERS and save ray-object intersection tests and

> assumptions about ray-object intersectons. This is not very applicable for
> anims, of course.
>
> Bruno.
Ah, OK, I misunderstood your idea, then. The logical conclusion is: Cool,
the idea I posted here is mine!!! ^_^
Regarding your idea, it was mentioned in the off-topic forum
(http://news.povray.org/povray.off-topic/thread/%3Cweb.471cbb5fa19b477d27eef15b0%40news.povray.org%3E/)
that in a raytracer by the name of Maxwell you can change lighting set-up
after rendering. This sounds very much like what you are suggesting, since
it requires storing somewhere previous render information. So it is
possible, but not in POV just yet. It's definitely the way to go for future
versions of POV-ray, IMHO. I may give it a go in the SDL raytracer,
actually, assuming I can figure out how it works exactly and if I can find
the time.

Post a reply to this message