Canon EOS R5 MK2 Optimised Noise Reduction Pre-sets.

[Josh-H]

A couple of days ago I finished work on the Noise reduction pre-sets for the R5MK2. A lot of work went into these and how they were optimised. For anyone who is interested: https://blog.jholko.com/2024/08/27/adobe-lightroom-iso-specific-noise-reduction-pre-sets-canon-eos-r5-mk2/.

[francois]

Thanks for the interesting article!

[Redcrown]

Thanks for that. I only did a quick read and it appears this work is only for the "old style" noise reduction and sharpening in Camera Raw. Have you done any work on the newer Denoise AI? I've tried to determine optimum settings for the Denoise AI single amount slider based on ISO and have not reached any conclusion yet.

In fact, it appears the Denoise amount should be based on shadow noise only, not just ISO. An ISO 6400 image with many deep shadows gets a higher amount than an ISO 6400 image with only midtones and highlights. And even then, it appears this first implementation of Adobe Denoise AI does not support presets.

[Josh-H]

In short, the presets are design to work either stand-alone or with the AI noise reduction option in Lightroom. The AI noise reduction is extremely processor intensive and very slow unless you have a very fast computer. The pre-sets are instantaneous on import so give you an excellent starting point. They are optimised to make a given ISO look as good as possible.  If you need additional Noise Reduction you can then apply the AI Noise Reduction on top. Or use a third party option like Topaz.

Typically, I have found with LR AI Noise Reduction a setting somewhere between 40 and 55 is optimal in most cases (assuming you are not doing any other noise reduction - including the pre-sets. However, your mileage may vary as there are so many factors that you need to consider. Was the exposure optimal? What ISO? Etc. Not every image needs it either - it depends on how noisy the file is. And if you are applying the noise presets and then applying AI Noise reduction on top then you will want less of the A.I noise reduction. How much? Again, it depends on the file. No easy answer on this.

Up to ISO 3200 with the EOS R5 MK2 the pre-set should be sufficient if the file was optimally exposed in the field. At ISO 6400 and above, you may want to add some small amount of additional noise reduction with the AI sliders.

Hope this helps

[Redcrown]

Are you sure about that?. I believe that when you "enhance" a raw image with Adobe Denoise AI, it ignores any luminance and color noise reduction settings that have been set manually or by preset. The resulting dng then has luminance and color NR set to zero. So, you can't apply Denoise AI "on top of" normal noise reduction. You can, of course, apply additional luminance NR to the dng created by Denoise AI, but I don't know why you would do that.

I appreciate your advice that Denoise AI amounts of 45 to 50 are optimal. That is my experience as well, but I've not seen it discussed anywhere yet.

[Josh-H]

Im not so sure about this statement that " I believe that when you "enhance" a raw image with Adobe Denoise AI, it ignores any luminance and color noise reduction settings that have been set manually or by preset. The resulting dng then has luminance and color NR set to zero. So, you can't apply Denoise AI "on top of" normal noise reduction. "

My understanding is that like any other adjustment you make to the RAW file it is preserved by the AI Noise algorithm when you run this. I will see if I can get some confirmation either way on this from Adobe and revert.

[Josh-H]

I am waiting on further clarification directly from Adobe. But according to their blog and in the meantime:
"As with previous Enhance features, any adjustments you made to the source photo will automatically be carried over to the enhanced DNG. You can edit this DNG just like any other raw photo, applying your favorite presets and custom tweaks."

Based on this statement, any changes you make to the noise reduction sliders are respected by AI noise reduction in LR.
Adobe Noise Blog Article

[mcbroomf]

I am interested in this discussion as I just zero out all noise reduction settings (and also sharpening) before I use Enhance Denoise.

Josh, I could not find any reference to noise reduction settings being preserved.  In fact they are zero'd out.

"What about the previous Noise Reduction sliders, like Luminance and Color? They’re still here, but they’re tucked away in a new sub-panel called Manual Noise Reduction. Note that Denoise will automatically set these sliders to zero on the new DNG."

It's easy to see this (ie that the NR sliders are set to zero in the resulting DNG).

It's also pretty easy to verify that nothing is being done (ie your NR settings are not baked in even though the sliders are set to zero).

1) Set NR sliders to zero and run Denoise
2) Adjust NR sliders to any value and run Denoise (same value as 1)
3) Open both DNG files into layers in PS and set them to Difference

You will see that the files are identical (in my test anyway).

Addendum : After testing as above and getting R:G:B values of 0 all across the difference of the layers I thought I'd check that the regular NR sliders do in fact make a measurable difference.  On one of the layers (both were opened as smart objects) I opened ACR and set both luminance and colour sliders to about 30-40.  Once I did that and the layer updated I could see many non-zero values across the image.  All low values, mostly single digits, a few at around 10.

[digitaldog]

1) Set NR sliders to zero and run Denoise
2) Adjust NR sliders to any value and run Denoise (same value as 1)
3) Open both DNG files into layers in PS and set them to Difference

You will see that the files are identical (in my test anyway).

Addendum : After testing as above and getting R:G:B values of 0 all across the difference of the layers I thought I'd check that the regular NR sliders do in fact make a measurable difference.  On one of the layers (both were opened as smart objects) I opened ACR and set both luminance and colour sliders to about 30-40.  Once I did that and the layer updated I could see many non-zero values across the image.  All low values, mostly single digits, a few at around 10.

That's the test to run, agreed. Tiny differences are to be expected, but if not much larger, they would bode well for your 'theory'.
This Apply Image tutorial covers how I do this kind of visual analysis for differences:
http://digitaldog.net/files/Apply_Image.pdf
The other way would be to do so colorimetrically, but these big docs would choke a product like ColorThink Pro, and we don't want to do any resampling in this kind of testing. The Apply Image should be sufficient.

[Josh-H]

I heard back from Eric at Adobe:

if you (manually or otherwise) adjust the old NR sliders prior to running Denoise, those slider values will be reset to zero on the new DNG, and the effect of those sliders will not be baked into the DNG.  In other words, the new DNG will have only AI Denoise applied, and no manual/legacy NR applied.  So, it's as if you're starting fresh.

The reason for this behavior is that AI Denoise was trained and designed to be used alone, without any of the manual/legacy NR applied.  This said, I recognize that in some cases, even with AI Denoise, there may be some residual noise that a photographer may wish to clean up manually afterwards, so the manual NR controls remain available to use "on top of" AI Denoise.

So that totally clarifies it.