Search found 42 matches

by Zelcious
Sun Oct 14, 2012 9:30 am
Forum: Links & Papers
Topic: Vertex merging paper and SmallVCM
Replies: 33
Views: 30638

Re: Vertex merging paper and SmallVCM

The last scene, with the two spheres and the blue and yellow walls, how are the lights in the roof constructed? The path tracing image looks really off to me but I guess there is a natural explanation to that. My best guess would be a lens in the middle not covering the entire hole in combination wi...
by Zelcious
Sat Oct 13, 2012 8:48 am
Forum: General Development
Topic: MLT Singularities
Replies: 7
Views: 4811

Re: MLT Singularities

It matters because you will spend a lot of time searching for relevant light sources that could be better spent actually rendering the image. I did a similar approach when trying to find light sources in a volume for vpls' and it paid off really bad. Brute force path tracing was much faster. Also I ...
by Zelcious
Tue Oct 09, 2012 3:54 pm
Forum: General Development
Topic: MLT Singularities
Replies: 7
Views: 4811

Re: MLT Singularities

Since you said caustic paths, have you tried anything (progressive) photon mapping related? Not in this project but in previous projects. However don't think it's applicable in this case since I have very large scenes and anything could be a light source. For it to work would need some kind of adap...
by Zelcious
Tue Oct 09, 2012 9:02 am
Forum: General Development
Topic: MLT Singularities
Replies: 7
Views: 4811

Re: MLT Singularities

Wenzel's manifold exploration looks really great but it will only help if it actually exists a nearby path (doesn't have to) and if the specular surfaces are smooth enough (geometrically not scattering wise). In my test scene I only had hard caustics paths, so really narrow effects could cause reall...
by Zelcious
Sun Oct 07, 2012 2:16 pm
Forum: General Development
Topic: MLT Singularities
Replies: 7
Views: 4811

Re: MLT Singularities

Is it ok to do importance sampling around high contribution samples and then use MIS on both of them? Or do you introduce bias? Seems like a really good strategy to me otherwise. Similar to MLT but with more freedom and less negative aspects.
Nevermind this stupid nonsens, it's obviously incorrect.
by Zelcious
Sun Oct 07, 2012 9:13 am
Forum: General Development
Topic: MLT Singularities
Replies: 7
Views: 4811

MLT Singularities

Are there any unbiased methods to avoid MLT based algorithms getting stuck in singularities. Or do you just have to forget about unbiasedness and set a limit on max consecutive rejections (Seems like an ugly hack to me) Importance sampled algorithms have a clear advantage here. Why aren't they used ...
by Zelcious
Fri Oct 05, 2012 8:12 am
Forum: General Development
Topic: The light image in BPT
Replies: 11
Views: 8006

Re: The light image in BPT

I find it much more intuitive if you consider the path between the image film and the lens: you sample the image film uniformly and the lens uniformly. You want to assign equal weight to each sample so you have to divide by the geometry term between any two points so what you have is: pA(x_film) * p...
by Zelcious
Fri Oct 05, 2012 8:06 am
Forum: General Development
Topic: question about woodcock tracking for volume transmittance
Replies: 1
Views: 2912

Re: question about woodcock tracking for volume transmittanc

Woodcock tracking is used to find the first interaction of a particle inside a volume. This distribution is equal to the transmittance. So for a single sample, you have pdf=1 and the transmittance is also 1 since there are no collision before this sample. Right? Transmittance between two points is t...
by Zelcious
Sat Sep 29, 2012 10:58 pm
Forum: General Development
Topic: The light image in BPT
Replies: 11
Views: 8006

Re: The light image in BPT

First of all, a philosophical question: Why are people so crazy about this projected solid angle measure of Veach's? Why do you need to bother with it, as it is unintuitive, and its only purpose is to (artificially) put the cosine in the pdf instead of the contribution function when doing MC integr...
by Zelcious
Fri Sep 28, 2012 11:13 pm
Forum: General Development
Topic: The light image in BPT
Replies: 11
Views: 8006

Re: The light image in BPT

So the term abs(dot(r.dir, n)); is to convert to projected pdf right? Now think about this: You spawn particles equally on the surface and equally in the directions. Equal radiance in all directions right? Draw a little surface with first a few particles pointing along the normal then a few particle...