Altus rendering denoising system

[voltaire585]

Rather than have Altus as a 3rd party plugin, i would really prefer an Otoy internal solution such as porting the Brigade denoiser, writing Otoys own , or using an adaptive denoiser similar to Fstorm. It would solve problems future compatibility problems with Altus and improve ease of use.

[Goldorak]

We have our own plans for eliminating noise within Octane in parallel to this plug-in. Brigade's denoiser was for real time use cases, it has to be adapted for final quality rendering as well.

[Rebelismo]

Depends on your source image I would say; after a decent amount of samples you shouldn't loose much.
Compare the renders I posted at the beginning.

From the Corona forums
Rendertime: 1:40 min (4 passes, which is almost nothing)
Filtering Time: 43Sec

image.jpeg

image.jpeg

Here's a short review:
https://youtu.be/BMhAVnKGPM4

Altus is looking nice, but I think that a lot more samples would be needed for this scene. Those light blotches are pretty extreme. Curious to see how the Brigade de-noise implementation will work in Octane.

[RobSteady]

Depends on your source image I would say; after a decent amount of samples you shouldn't loose much.
Compare the renders I posted at the beginning.

From the Corona forums
Rendertime: 1:40 min (4 passes, which is almost nothing)
Filtering Time: 43Sec

image.jpeg

image.jpeg

Here's a short review:
https://youtu.be/BMhAVnKGPM4

Altus is looking nice, but I think that a lot more samples would be needed for this scene. Those light blotches are pretty extreme.

Obviously the render time is too short, that's why I wrote:
Rendertime: 1:40 min (4 passes, which is almost nothing)
But even for this grainy input image the result is pretty impressive. Complaining in this case is just ridiculous.
We have complex interiors not cleaning up after +6000 samples with high frequency noise; for this it would be the perfect solution.

[Rebelismo]

Depends on your source image I would say; after a decent amount of samples you shouldn't loose much.
Compare the renders I posted at the beginning.

From the Corona forums
Rendertime: 1:40 min (4 passes, which is almost nothing)
Filtering Time: 43Sec

image.jpeg

image.jpeg

Here's a short review:
https://youtu.be/BMhAVnKGPM4

Altus is looking nice, but I think that a lot more samples would be needed for this scene. Those light blotches are pretty extreme.

Obviously the render time is too short, that's why I wrote:
Rendertime: 1:40 min (4 passes, which is almost nothing)
But even for this grainy input image the result is pretty impressive. Complaining in this case is just ridiculous.
We have complex interiors not cleaning up after +6000 samples with high frequency noise; for this it would be the perfect solution.

It's not about complaining, it's simply about standards. I'm going to be using Altus with Octane as well, but that particular image is just not a good demonstration. I understand that you have to crank up the samples to get something better, but don't forget that altus needs dual inputs to denoise. I think a much better demo would take into consideration the min samples it takes for something to look good, plus the time it takes to render that frame again.

[RobSteady]

Don't care if it's Altus or something else but can we please get something like this already?!

Otoy, wake up!
https://corona-renderer.com/blog/corona ... #more-2056
http://www.recentspaces.com/blog/2016/4 ... omparisons

[RobSteady]

Another one passing by. Octane stumbling, lagging behind:
http://cgpress.org/archives/denoising-t ... y-way.html
https://www.linkedin.com/pulse/denoisin ... ele-lecchi


(*) V-Ray can update the denoised result during progressive rendering so you can check out how it’s coming along and stop the render if the denoised result is good enough;
(*) The denoiser has a mode to operate on different render elements and then compose the final image back; this variant allows it to better preserve texture detail;
(*) You can go back and adjust the denoised result after a render is complete to try out different settings;
(*) The V-Ray denoiser supports hardware acceleration which can make it anywhere between 5 and 15 times faster than the pure CPU version;
(*) There is also a standalone command-line denoiser that can be used to denoise images outside of 3ds Max. This is useful in several situations:
a) when denoising animations, the denoiser can look at several frames at once. This allows you to get better results with noisier original frames and lower render times overall;
b) if you need to do some kind of post-processing, like put together one image stitched together from many tiles, you can run the denoiser on the final stitched result.
c) you can denoise images coming out from other render engines.

[Goldorak]

Another one passing by. Octane stumbling, lagging behind:
http://cgpress.org/archives/denoising-t ... y-way.html
https://www.linkedin.com/pulse/denoisin ... ele-lecchi


(*) V-Ray can update the denoised result during progressive rendering so you can check out how it’s coming along and stop the render if the denoised result is good enough;
(*) The denoiser has a mode to operate on different render elements and then compose the final image back; this variant allows it to better preserve texture detail;
(*) You can go back and adjust the denoised result after a render is complete to try out different settings;
(*) The V-Ray denoiser supports hardware acceleration which can make it anywhere between 5 and 15 times faster than the pure CPU version;
(*) There is also a standalone command-line denoiser that can be used to denoise images outside of 3ds Max. This is useful in several situations:
a) when denoising animations, the denoiser can look at several frames at once. This allows you to get better results with noisier original frames and lower render times overall;
b) if you need to do some kind of post-processing, like put together one image stitched together from many tiles, you can run the denoiser on the final stitched result.
c) you can denoise images coming out from other render engines.

I see nothing on this list that the integrated Altus plug-in we showed in April doesn't already cover.

3.1 allows any 3rd party denoiser to integrate live (or offline) in Octane via the 3.1 native plug-in API, of which Altus is the first to leverage (along with incoming post and imager plug-ins e.g. OctaneImager). In parallel to this, we are working on reducing noise in the engine itself before even getting to a denoiser step.

[RobSteady]

3.1 allows any 3rd party denoiser to integrate live (or offline) in Octane via the 3.1 native plug-in API, of which Altus is the first to leverage (along with incoming post and imager plug-ins e.g. OctaneImager). In parallel to this, we are working on reducing noise in the engine itself before even getting to a denoiser step.

Please give us schedule when all this is going to happen.

[RobSteady]

tick tock tick tock...
https://www.youtube.com/watch?v=HMDrTS1aEPk