SPPM renderer (CPU-only)

[guibou]

I wonder if there is a way to combine SSPM's caustic ability with plain bidirectional path traicng for all the other regular conditions

This is what everybody think since the first SPPM paper was presented.

There was a nice proposition at siggraph asia 2011 https://graphics.cg.uni-saarland.de/201 ... x-merging/

Toshiya published a technical report with such results recently :
http://cs.au.dk/~toshiya/ (scroll to technicals report)

I was working on something similar until this report was published. The crappy thing in research is that generally people have the same ideas, but some of them are really quicker to get result and publish.

<sad mood> /me is looking for a new job

[Dade]

There was a nice proposition at siggraph asia 2011 https://graphics.cg.uni-saarland.de/201 ... x-merging/

Toshiya published a technical report with such results recently :
http://cs.au.dk/~toshiya/ (scroll to technicals report)

God .. they look like the missing link between human being and the divinity

It is a while this idea of merging path vertex on BiDir is on the "air". It looks like it is time for aiming for the definitive rendering algorithm Luckily, the definitive algorithm is never going to exist but this starts to look like a solid foundation.

[SATtva]

I can haz cheezburger it in lux pliz?

[CoolColJ]

I am finding SPPM not so good at outdoor type scenes? Like you might need a lot more photons...

[ingenious]

I am finding SPPM not so good at outdoor type scenes? Like you might need a lot more photons...

This is fixed automatically, as much as possible, by the combined vertex merging and bidirectional path tracing algorithm discussed above. I think Lux will have a new bidirectional progressive algorithm a bit later this year It doesn't solve all difficult cases (e.g. combinations of highly glossy and specular materials), but it's definitely more robust than either bidirectional path tracing and (S)PPM alone.

[Dade]

I think Lux will have a new bidirectional progressive algorithm a bit later this year

Uh, uh

[CoolColJ]

SPPM murders MLT in this test - the underwater caustics and it's reflection in the sphere rendered within 1 min! 15 min render for both
But the pillars are more coarse. And yet in other times, SPPM can be a slow splotchy mess

SPPM - 2million photons, 3 start radius 0.5 arc. filtering is off

MLT

Now if you turn off direct light sampling in SPPM - the smoothing in the pillars doesn't work?
15 min rendering as well - smooth caustics...

[Meelis]

Why would you prefer reall geometry to smoothing?
reall geometry looks much better.

[guibou]

SPPM murders MLT in this test - the underwater caustics and it's reflection in the sphere rendered within 1 min! 15 min render for both
But the pillars are more coarse. And yet in other times, SPPM can be a slow splotchy mess

It is the reason why research is looking for a merge between the twos methods.

Now if you turn off direct light sampling in SPPM - the smoothing in the pillars doesn't work?

Hoo, no, not that one again... I was thinking we where done with it...

--
Guillaume

[SATtva]

Now if you turn off direct light sampling in SPPM - the smoothing in the pillars doesn't work?

Hoo, no, not that one again... I was thinking we where done with it...

Direct lighting was disabled in the last test. Isn't that an expected behavior?