Unbiased Randering and Caustics

[vlado]

Another classic:



Best regards,
Vlado

[jacksc02]

WOW!!! that looks fantastic!

[anchovy]

Vlado-

Ok im im super excited by these, not sure when ill get to play but im going to be counting the days!!! looks like you are getting some nice results. Id love to hand you a couple of the scene files I had cooked up when we were first talking about this if you want to use them for some tests. Id also be interested to see some times. Certainly its resolving the reflected caustics much better. Any thoughts to if you are sacrificing much for it in the way of introducing other noise?
Also do you feel that the overall image is resolving faster or just differently?

Thanks for sharing!

-Michael

[vlado]

Yes, feel free to send over any test files that you have They will be very helpful. You can send them to vlado@chaosgroup.com

Best regards,
Vlado

[Dubbie99]

I have got a large scene that I am very eager to try.
Please let us know as soon as you merge into the nightlies.

[anchovy]

Vlado-

I sent you a handful of files to play with. Cant wait to see some results.

-Michael

p.s. I love to pool caustics!

[deflix]

hands up who's excited about vray 3.0?

[deflix]

This is patented last time I looked into it. So forget about it

Best regards,
Vlado

sad....I regret filing one years ago myself.

“The idea that I can be presented with a problem, set out to logically solve it with the tools at hand, and wind up with a program that could not be legally used because someone else followed the same logical steps some years ago and filed for a patent on it is horrifying.”

― John Carmack

[anchovy]

Vlado-

I hadn't seen this paper. I might be of interest though at this point im not sure there are too many loose ends for your implementation.

http://www.iliyan.com/publications/VertexMerging

-Michael

[vlado]

Yep, I've seen it. I know Iliyan personally too.

Best regards,
Vlado

[^Lele^]

MLT doesn't attempt to find anything; it relies on blind chance to find important parts (same as any other path tracing algorithm), and then it explores nearby similar paths. But, same as other Monte Carlo algorithms, it can completely miss an important part of the image if it never comes across it in the first place.

Further on, MLT algorithms must use pure random numbers; it is not possible to use optimizations like Schlick sampling or Quasi-Monte Carlo sampling. This immediately makes them worse than regular (bidirectional) path tracing for a vast majority of not-so-difficult to compute scenes.

Finally, because of its pure randomness, using MLT for animations is difficult (there is no guaranteed consistency between successive frames).

Best regards,
Vlado

I was toying with the Mitsuba renderer, which sports just such a manifold exploration technique, and yeah, Vlado is ofc right.
While i'll admit it's the funkiest render method i have ever witnessed (with buckets spreading light in the scene before they hit the corresponding raster space of it, leading to progressive brightening of the whole frame from early buckets), the VERY chancy way it uses to determine paths means that unless there's an awful lot of very detailed paths, convergence will change quite dramatically across renders/frames/slight viewpoint changes.
Still extremely funky to play with.

[anchovy]

Yep, I've seen it. I know Iliyan personally too.

Best regards,
Vlado

Some nice results, Are you looking to his work to influence how VRay will be handling the vertex merging? Also Im surprised that their 24hour reference images are so clean. Not sure how those are being approached but they are much cleaner than I would have guessed.

-Michael

[anchovy]

Vlado-

Here is yet another little test. Ive resurrected the old winosi scene and set it up to somewhat match a version ive seen in one of the vertex merging papers



The interesting thing about this one is both of the lights are area spheres inside a reflector with a glass lens so every directly lit surface is I guess by definition a caustic.





Here ive removed a lens and you can see the illuminated reflector. Interesting that you dont see this through the lens in the versions where the lens is in place.

If anyone wants to play http://morphographic.com/PathTrace/w...orball_004.max

Thanks

-Michael

[anchovy]

Veach MLT Room

For fun here is my take on the Veach mlt room setup. Simple for sure but its a nice test all the same. I let this one run today while I was at the office. It was fairly clean within a couple hours. I like the subtle caustics behind the light on the right wall.



Here is the 2012 scene file http://morphographic.com/PathTrace/2...achRoom_01.max

-Michael

[vlado]

For fun here is my take on the Veach mlt room setup.

Do you have the scene for this too?

Best regards,
Vlado