[luis.hijarrubia]

Ok for images
And animations ?

Using object ID channel you can automatize things on animations. I know it will be easier if we can have a custom alpha with blending factor for denoising, and I will push to develop that, but for the moment you can work like that.

[Mihai]

For me what should be first on the agenda is giving us more control. And especially where it is needed the most - in Maxwell, so we can try different settings on the already rendered image. It seems the defaults used internally now are very heavy and we lose too much detail, especially medium tone detail (very bright and very dark details are more intact).

It would be great if someone from Altus could give us some tips. Their help page has minimalistic info to say the least.
https://support.innobright.com/kb/faq.php?id=17
-------------------------

Application options:
* --config= use a configuration file to specify program inputs
* -r, --radius= the filter radius (default 10)
* --kc_1= filter parameter that is sensitive to image detail, but less aggressive towards noise (default 0.45)
* --kc_2= filter parameter that has the same function as kc_1 but is used for the second pass (default 0.45)
* --kc_3= filter parameter that determines the aggressiveness towards noise (may cause loss of detail) (default 1x10)
* --kc_4= filter parameter that controls the sensitivity of the second pass filter to color differences. A higher value leads to more agressive filtering. (default 0.45)
* --kf = filter parameter that controls the sensitivity of the all candidates, and the second pass filter, to feature AOVs

* --quality= filter the image at different quality levels, {preview, production}

--------------------------

* --kc_1= filter parameter that is sensitive to image detail, but less aggressive towards noise (default 0.45)

So if I want it to be less aggressive towards noise, what should I do, increase it? Decrease it?

* --kc_3= filter parameter that determines the aggressiveness towards noise (may cause loss of detail) (default 1x10)

Again, what should I do to make it retain more detail? And what is this parameter unit, 1x10? If I want to change it, should I use 2x10, 3x10, 4x10......or 1x11, 1x12, 1x13.....?

[luis.hijarrubia]

What i understand is that KC1 keeps detail, so bigger value more detail, less noise reduction. KC2 the same for second pass.
KC3 is opposite, the bigger more noise reduction, less detail. And in the standalone app it's a float number from 0.0 to 5.0.

[Eric Nixon 20170610100005]

What do you think about the AI based denoiser in IRAY which amd or nvidia demo'd recently. It appears to retain details thru glass.