Denoiser and Refract affect all channels

[Morne]

When you have refractive objects that affect all channels, the denoiser results in BRIGHTER parts of the image where the refraction bits are.

For example a balustrade with objects behind it need affect all channels otherwise the multimatte objects behind the glass doesn't work. But now the denoiser brightens the balustrade

[svetlozar.draganov]

I did a quick test but didn't get difference in brightness between the RGB and the denoised result except the slightly jagged edge of the refractive object(not sure if that can been seen on the screenshots bellow).


Is it possible to send us a scene that reproduces the issue for investigation?

[Morne]

Almost 2 months ago. Done and dusted. Will let you know on the next one

[svetlozar.draganov]

Sorry for the very long delay, there was enormous amount of requests and some threads were pushed back.

[Morne]

No worries. That one wasn't very critical and it was easy to fix in post.

[olly]

We are having trouble with this - was there ever a solution?

When you set the denoiser to 'denoise render elements separately' any refractive materials that are set to 'all channels' get brighter. Yet you need 'all channels' to see multimattes through the refractions...


RGB


Comped from elements or denoiser


Multimatte

Max file: refraction 'color only' vs 'color+alpha' vs 'all channels'.zip

Please let me know if I am missing something obvious here!

Also, could someone advise on how to include the denoised render elements in the standalone denoiser's output?

Many thanks,
Olly

[vlado]

Thanks for the file, will look into it. There's no way to save the denoised elements back right now, but it's on the "to do" list. If you use Nuke, you can use the Nuke denoiser plugin - in that case you can replace the input RGB element with any other element you want to denoise.

Best regards,
Vlado

[olly]

vlado firstly thank you for adding the ability to denoise render elements in vray 4 - works very well.

Unfortunately these bright refractions are still present in denoised renders in vray 4...

We would really love to see this fixed - it is a quite a disruptive bug in our workflow!

Cheers,
Olly

[vlado]

Unfortunately these bright refractions are still present in denoised renders in vray 4...

I don't know how this would be possible - the "denoise render elements separately" method is completely removed from the denoiser code. Or are you talking about the comp part? Maybe you can give more specific steps to reproduce the issue with V-Ray Next?

Best regards,
Vlado

[olly]

Hi Vlado,

Yes the comp part. If a and RGB is reconstructed out of the component elements (either manually in Photoshop or by the denoiser with -elements=1 ) refractions get brighter. Not sure if it helps, but this only happens to materials with "affect all channels" set for refractions in VRayMtls.

The scene supplied above demos this.

Hope it can be fixed.

Cheers,
Olly