Corona Denoiser Run Time

[innobright]

Hello,
     I am curious if anyone has some benchmarks for Corona denoiser's render time for a given size or frame resolution.   Like how long did the render take with denoiser turned off (higher quality settings) and how long did the same scene take with denoiser turned on (lower quality settings)? 

Please help.

Thank you.

[disoranno94]

I'm a recent user and I'm also interested in this. Usually I have the denoiser turned on, but I'm considering turning it off and comparing the result.

[PROH]

Hi. The "Denoise Amount" value, is not a quality parameter, but is a mix-value, setting the blend between a full denoised image (1) and a full "normal" image without denoising (0).

So to compare ON/OFF you'll have to set "Denoise Mode" to "Full denoising", and then change the amount value between 1 and 0 (0 = OFF, 1 = ON).

I can't give you any benchmark-numbers, but you can easily set up your own test.

Hope it helps

[romullus]

You can run tests if you wish, but for me it's very obvious, that denoising takes only fraction of time, that would be needed to achieve comparable results without denoising. The only shortcomings of denoising, that i can think of, is slight loss of subtle details (though easily adjustable with denoise amount at any time) and higher demand for system RAM. Usually its usage is no brainer for me - if i know that my scene won't occupy all available RAM, i turn denoising on.

[TomG]

And for tests from the real world, in the Recent Spaces interviews they discuss how they saw render times reduced by 50 to 75%:
https://youtu.be/TwvFW4zBv14?t=6m34s

[maru]

There are some comparisons here:
https://coronarenderer.freshdesk.com/support/solutions/articles/12000004976

The "denoise amount" is also interactive, meaning that you can change its value during/after rendering and see how the image with and without denoising compares.

[SharpEars]

On the topic of de-noising, Vlado just gave a nice long detailed technical talk on renderer denoising techniques in general and the implementation of denoising in V-Ray, in particular. The theory made a lot of sense and the quality of the results presented definitely helped elucidate and exemplify the theory.

I am wondering if Corona uses similar techniques to the "bleeding edge" ones he discussed or some of the more primitive ones at the start of the presentation.

Link to the talk: https://www.youtube.com/watch?v=UrOtyaf4Zx8

[Nejc Kilar]

Thanks for the link SharpEars! You do have some sharp ears there buddy...