stainless steel in low light

[tricky]

I often come up against an excessively noisy render of stainless steel when working on low-light, atmospheric interior shots. For external stainless steel clad buildings, I can usually get pretty nice, smooth results, but there is usually plenty of light coming from the vraysun and sky. Intenally, where I want the bare minimum of 'task lighting', the stainless steel generally looks awful.

Any tips on how to get smooth (relatively) noise free renders of stainless steel objects under low light conditions?

My material settings are a mid-dark grey for diffuse, near-white (about 80%) in the reflection, glossiness set to anywhere between 0.6 and 0.85, subdivs set to 20. I use LC/Irmap.

[tct70]

This might be worth a look...........

http://www.vray.info/topics/t0112.asp

[tricky]

That doesn't really help me. In good lighting (ie when there is plenty of light in the scene) that would work fine, but I have issues in low light images. I'll post a sample when I get a chance.

[crazy homeless guy]

more subdivisions?

[dejaVu]

Hi Tricky,
Your problem with noise SS under low lighting conditions might have to do with the image sampler not doing enough samples on the glossy reflections because when it's dark, the contrast between pixels probably fall under the qmc noise threshold.
Therefore, the sampler views the neighbouring pixels as smooth enough and does do anymore sampling.
You could try reducing the threshold, or using a VRayPhysicalCam increasing exposure so that the image is brighter without changing your lighting setup.

[werticus]

i've been having the same problem... and sub divs (on the mat) does not seem to help nor does noise threshold

the only thing i could get to work was global subdivisions!

so i did LC and saved it and i did IR and saved it and then turned up the global subdivs for the final render and sent it off.

[paulison]

wow very nice description of what causes this Vu.

We have this problem a lot on shadowed sides on buildings on rollerdoors for example with fine geometry - there is lots of blotching. Generally we end up adding more light

[tricky]

But adding more light (or changing exposures) is not a good solution as we want to achieve the atmospheric darker look.

Any other ideas?

[studioDIM]

if lighting is dimmer, so will likely be the difference between adjacent pixels.
My guess would be a smaller noise threshold might help, possibly coupled with a 2.2 gamma with "adaptation only" on (will apply gamma only to the sampling, possibly stretching a bit the differences between pixels, making it an easier job for the QMC sampler to pick them out as good candidates for supersampling).

A file to play with would be mighty, as usual.

Lele

[flipside]

Yeah lower noise th should do it, maybe also try lowering the adaptive amount (default 0.85). Or brighten it up and darken again in post

[sv]

possibly coupled with a 2.2 gamma with "adaptation only" on

yeah - i was going to say this also.

But thinking about it further - could you say set the gamma to something really high like 4/5 or something; would this theoretically make it even easier for the QMC sampler and thus make a cleaner rendering?

Just thinking outloud.....?

[studioDIM]

my guess is that then the AAser would have to work overtime as adjacent pixels might be very far apart...

Might be tried...

Lele

[timmatron]

Which AA are you using? You should use adaptive subdivision and not adaptive qmc, and you will get a lot less noise on the ss. Also use just the default area filter and not any sharpening filters. You can always sharpen images up in photoshop.

[flipside]

Adaptive subdivision smooths out the noise, it doesn't get rid of it imo. Instead you get a more blotchy noise while you get sharp noise with adaptive qmc. Adapt QMC is also a lot quicker with glossies...