Back to Beauty Compositing -> Denoiser not working

[thescopeteam]

Hello, we are investigating the possibility of the vray denoiser. What ever we try it does not work. This is an example from Photoshop with a test scene from you. In Nuke its the same result btw. All render elements from Back to Beauty does not result in the same image like the main beauty RGBA (16bit float). Normally we use the back to bty workflow in conjunction with light selects. With light selects it does not work at all. All RE have the option "denoiser "activated in this Photoshop example.
I can send you the scene I used (40mb) I used the vray denoiser with mild preset or default. I tried to denoise the image with the vray denosier tool for animations. Nothing really works. This is how it supposed to work. Does is work this way, are we doing something wrong, bug ?

http://www.grapheastegypt.com/vray/b...ion/index.html
https://docs.chaos.com/display/VMAYA/RGB_Color

Vray 5.20.02, OCIO ACES

back to beauty




difference between back to beauty and RGBA




subtracted light selects from back to beauty



back to beauty layer all "Linear Doge Add" except the bottom one






thanks and cheers Florian

[hermit.crab]

Are you certain you're comparing the denoised RGB with the composed denoised elements? From what I see, you compare with the RGBA, not the effectsResult / Denoiser channels. Otherwise, I tested on a simple setup and everything works as expected.

[thescopeteam]

Hello Aleksandar, we tested it again. I uploaded the psd files and Maya file to DropBox. The main purpose is actually to render light selects and subtract them from a "back to beauty". It does not match, no matter if the render elements are denoised or not. Denoised it looks even worse. Are we doing something wrong? I used the Vay Denoiser with preset Mild. Different presets result in even more differences. I tested the denoiser tool as well, always with the same result
thank you Florian

V-Ray 5 for Maya, update 2.2
(v5.20.02 from Apr 27 2022)

[hermit.crab]

We've already worked on this issue (internal bug-tracker id: VMAYA-9768 ), however, it seems it's not fixed for all cases. Subtracting the VRaySun from the Beauty still yields negative pixels. I'll notify the project managers about it.