Comparing Noise Reduction for images (Raw)

[Hali]

I went to a macro-abstract (or was that an abstract-macro) workshop with my camera club today. The weather was poor and they were cutting the grass outside so we had to do everything inside and the light was pretty poor. Since I was shooting at high ISO I thought this was a good time to compare the big denoise contenders.
Methodology: The original Raw file was from a Canon R5 imported into LR with a Neutral Adobe Preset, Lens corrections applied, sharpening set to default. The ISO was 12800 1/250 sec f/5.0 (I was hand holding and even with the wonderful stability of the camera and the lens my left hand tremor requires some pretty good shutter speed to get crisp images.) I went to 100% in Lightroom and then ran the same image through 1. LR AI Denoise (new release) 2. DxO Pure Raw 3 3. Topaz Denoise - Raw 4. Topaz Denoise - Low light. 5. Nik Dfine 2. Here are the 5 with the originalplus the original

I think that DxO did the best job, although there is a slight color shift. The new LR/Camera Raw AI denoise was second best followed by Topaz Denoise for low light and finally Topaz Denoise Raw and the the Nik Define
. What are your thoughts?

 

[JustUs7]

I’ve downloaded a couple trial versions of the ones people swear by. I think I prefer a little noise and a more natural look. Whatever Canon does by default is fine by me. And now with the R6II I’ve got no complaints.

Topaz did this thing where it would remove all built in noise reduction before it applies their noise reduction. I was really impressed by the difference between the imported file and the denoise result. Then I went back the original file in DPP4 and noticed there really wasn’t much of a difference.

Heavy handed noise reduction still looks smeary to me. In images like the samples above. Smearing doesn’t matter because the colors are all blocks anyway. Throw in some hair or feathers and then judge.

 

[Tui]

I've only tried DXO Pure Raw and the new Camera Raw Denoise (Enhanced).
This is the original Raw file

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

These 2 below are the Raw file with shadows lifted

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

These 2 are ACR Denoise - some strange colour happening

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

These 2 below are DXO Pure Raw

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

This finished using DXO Pure Raw, still noisy but I don't mind it

• Canon EOS R6
• RF16mm F2.8 STM
• 16.0 mm
• ƒ/3.2
• 1/100 sec
• ISO 6400

 

[MikeZ]

Regarding the DxO results - I use PhotoLab 6.5 which includes Pure Raw 3.3. When processing raw files, DxO automatically applies lens and camera "corrections" which may cause the color shift. Further, there is a slider that enhances/increases the color space (or so DxO claim). I've not used PureRaw as a separate tool, so I don't know if any of those PL6.5 features are included and may be the source of the shift.

 

[Hali]

Regarding the DxO results - I use PhotoLab 6.5 which includes Pure Raw 3.3. When processing raw files, DxO automatically applies lens and camera "corrections" which may cause the color shift. Further, there is a slider that enhances/increases the color space (or so DxO claim). I've not used PureRaw as a separate tool, so I don't know if any of those PL6.5 features are included and may be the source of the shift.

I sometimes use PhotoLab 6.5 as well and the noise reduction in pure raw 3.3 has more options than the stand alone pure raw 3. I think they don't want to give away everything with the stand alone. I'm guessing we will see it in pure raw 4 when they have something to add to their photolab 7 software.

 

[Tui]

Regarding the DxO results - I use PhotoLab 6.5 which includes Pure Raw 3.3. When processing raw files, DxO automatically applies lens and camera "corrections" which may cause the color shift. Further, there is a slider that enhances/increases the color space (or so DxO claim). I've not used PureRaw as a separate tool, so I don't know if any of those PL6.5 features are included and may be the source of the shift.

In the examples above the colour shift was only in the Denoise feature of Camera Raw, the DXO Pure Raw did not affect the colour. I got the Pure Raw in 2021 and haven’t upgraded since then so it’s probably the basic one, it doesn’t have many options like 3.3

 

[Focus101]

For lower ISO, 6,400 to 10,000, ACR denoise enhance did a pretty good job for me. DXO pure Raw 2 is my go to Denoise anyways. I read the Adobe Blog - Denoise demystified.
"As a rule of thumb, we wanted to deliver clean, usable results for a 20 megapixel full-frame camera at ISO 51200. That’s a high bar!". If this was created and tested for 20MP files, what kind of results should we be expecting for R5 45MP? That is more than twice the MP.

 

[cdgta]

I just want to comment on DxO to point out that the result you will get depends on which level you use (ie Prime, Deep Prime, DeepPrimeXD). Many people seem to find that DeepPrimeXD works best on certain images while Prime or DeepPrime are preferred on others so you may want to run your own tests to compare these too.

 

[Only RF]

I have nothing to compare since I dumped all my 3rd party apps last week but I can show what I'm getting . I'm not too concerned if DXO or even Topaz are better products because Adobe Denoise is good enough for me. I've waited for many years for this and I know it is just going to get better. For the first release I'm pretty impressed.

ISO 40,000 from our trip to Japan. 1st file is the original. The 2nd file is after Denoise with the slider at the default of 50. I'm sure impressed with the detail on the door handles. 3rd file. The sky still had a little noise so I opened masking, masked it and reduced texture, sharpening and noise. 4th is the final.

 

[Only RF]

The doors might look a tad plasticky and since I can fix the sky I did another run with the Denoise slider at 35. This left a little noise which I'm OK with. It's all about personal preference. Maybe somewhere between 35 and 50. I know some you tube presenters liked about 43.