Respectable Denoise without DeepPRIME

I have an E-M5 Mark III camera. I have been developing RAWs using DxO PhotoLab 5 Elite Recently, I would like to test the denoise performance of various RAW developers.

You may get the original ORF file here

DxO with DeepPRIME is no doubt king.

However I found that the free OM Workspace with AI Noise Reduction can produce decent result. Then I feed the output TIF to On1 NoNoise AI. The final image is like this:

OM AI Noise Reduction + On1 Photo RAW (NoNoise AI)

In terms of noise, this one is respectable.

Your thoughts?

Annie Ng wrote:

I have an E-M5 Mark III camera. I have been developing RAWs using DxO PhotoLab 5 Elite Recently, I would like to test the denoise performance of various RAW developers.

You may get the original ORF file here

DxO with DeepPRIME is no doubt king.

However I found that the free OM Workspace with AI Noise Reduction can produce decent result. Then I feed the output TIF to On1 NoNoise AI. The final image is like this:

OM AI Noise Reduction + On1 Photo RAW (NoNoise AI)

In terms of noise, this one is respectable.

Your thoughts?

It looks fine to me, not especially noisy to begin, though. I would normally use Topaz DeNoise here for a little cleanup and sharpening, but this looks good with just a couple minutes in plain old Lightroom too...

Lightroom

Annie Ng wrote:

I have an E-M5 Mark III camera. I have been developing RAWs using DxO PhotoLab 5 Elite Recently, I would like to test the denoise performance of various RAW developers.

You may get the original ORF file here

DxO with DeepPRIME is no doubt king.

However I found that the free OM Workspace with AI Noise Reduction can produce decent result. Then I feed the output TIF to On1 NoNoise AI. The final image is like this:

In terms of noise, this one is respectable.

Your thoughts?

IMHO it is good - everything can be probably better, but that is highly individual. Agree with Erik - I will probably use TOPAZ instead of ON1 NoNoise AI. In some tricky situations it has just more controls over the process. Working with your file I tried to get rid of sometimes nervous noise pattern on the wall, soften bit mushy OOF parts, pick-up more details from white section of the fur and added some micro-contrast (superfine sharpening).  You can compare. How does it look to your eyes?

Erik Baumgartner wrote:

It looks fine to me, not especially noisy to begin, though. I would normally use Topaz DeNoise here for a little cleanup and sharpening, but this looks good with just a couple minutes in plain old Lightroom too...

May I know the reason DeNoise AI is better? Is it because we can choose models and strength of DeNoise?

I also have Topaz DeNoise/Sharpen/GigaPixel, but found that I need to run them one-by-one in order to get great result, while On1 Photo RAW is quite well-integrated, and includes AI Auto tunings on color/wb etc.

I have no plans to argue which one is better, just hope to get the better result of image retouching with strengths of different programs 

Kameratrollet wrote:

darktable + Free neural network https://github.com/trougnouf/mthesis-denoise

Some artifacts in the background. The network is not trained with Olympus cameras. Adding a little bit noise from the original image takes it out.

This is impressive for a completely free app.

_vlad wrote:

IMHO it is good - everything can be probably better, but that is highly individual. Agree with Erik - I will probably use TOPAZ instead of ON1 NoNoise AI. In some tricky situations it has just more controls over the process. Working with your file I tried to get rid of sometimes nervous noise pattern on the wall, soften bit mushy OOF parts, pick-up more details from white section of the fur and added some micro-contrast (superfine sharpening). You can compare. How does it look to your eyes?

I would love to learn how to control different parts of DeNoise AI. It seems I only use part of the program by just denoising the whole image with trials of different models and strengths of denoise.

Annie Ng wrote:

Erik Baumgartner wrote:

It looks fine to me, not especially noisy to begin, though. I would normally use Topaz DeNoise here for a little cleanup and sharpening, but this looks good with just a couple minutes in plain old Lightroom too...

May I know the reason DeNoise AI is better? Is it because we can choose models and strength of DeNoise?

I also have Topaz DeNoise/Sharpen/GigaPixel, but found that I need to run them one-by-one in order to get great result, while On1 Photo RAW is quite well-integrated, and includes AI Auto tunings on color/wb etc.

I have no plans to argue which one is better, just hope to get the better result of image retouching with strengths of different programs

I have other Topaz apps as well, but find DeNoise more than adequate for both noise reduction and sharpening when needed. I like that it has lots of fine tuning options and isn't nearly as slow as the others. Here's how I would handle this one with Topaz DeNoise and a small WB adjustment in the mix...

Lightroom/Topaz DeNoise

Annie Ng wrote:

_vlad wrote:

IMHO it is good - everything can be probably better, but that is highly individual. Agree with Erik - I will probably use TOPAZ instead of ON1 NoNoise AI. In some tricky situations it has just more controls over the process. Working with your file I tried to get rid of sometimes nervous noise pattern on the wall, soften bit mushy OOF parts, pick-up more details from white section of the fur and added some micro-contrast (superfine sharpening). You can compare. How does it look to your eyes?

I would love to learn how to control different parts of DeNoise AI. It seems I only use part of the program by just denoising the whole image with trials of different models and strengths of denoise.

If I understand properly (not sure) then you can process different parts of image quite easily and most of all - non-destructively. I assume you use ON1 PhotoRaw - that is assumption. Just go to layers and make two or three copies of the one (unprocessed) you see. You can e.g. totally denoise and cancel sharpening on the bottom one, then switch to the next above and create different AiDenoising approach on this one (e.g. leaving here some noise and adding detail enhancement). Then just switch the mask of this layer (upper one) to black (hiding everything - so you see only the bottom one) and paint a brush mask over the "top-layer-mask" to reveal the different AI approach of top layer (no need to be precise).

Obviously some work is necessary but in the final end - you can easily change the corresponding layer´s Denoising approach later - it is non destructive and re-editable. Topaz is not re-editable in this way. Once you do not like the result you did, you have to go back and reedit denoising from scratch unlike ON1 where you just fine tweak.

I hope I do not repeat the obvious.

sybersitizen wrote:

ON1 NoNoise AI is not just respectable - it's sometimes better than DeepPRIME, and it's faster, and it offers more user control, and it works with non-RAW files. The choice of what's 'king' comes down to the camera, the lens, the subject, and the shooting conditions. I use them both, and choosing based on any one image is not the way to go.

I discovered something similar recently.

If I tried to denoise a RAW image with PL5 DeepPRIME, then pass the DNG to Photo RAW, NoNoise AI still reduces the remaining noise after DeepPRIME. It seems to me that DeepPRIME and NoNoise target different types of noise. I have an image that is very obvious when denoised with both tools, however due to privacy issue I cannot post it online.

I wondered. "What happens when NoNoise AI is applied with another free AI noise reduction tool?" So I tried OM Workspace, and they showed a pretty significant improvement over the original image.

While the original image is not very noisy, it is already at ISO 6400, the highest ISO my E-M5.3 can use.

This combination is low cost (because OM Workspace is free, and Photo RAW is on sale), and the result is quite interesting to say the least. I am looking forward to the new 2022.5 release.

_vlad wrote:

If I understand properly (not sure) then you can process different parts of image quite easily and most of all - non-destructively. I assume you use ON1 PhotoRaw - that is assumption. Just go to layers and make two or three copies of the one (unprocessed) you see. You can e.g. totally denoise and cancel sharpening on the bottom one, then switch to the next above and create different AiDenoising approach on this one (e.g. leaving here some noise and adding detail enhancement). Then just switch the mask of this layer (upper one) to black (hiding everything - so you see only the bottom one) and paint a brush mask over the "top-layer-mask" to reveal the different AI approach of top layer (no need to be precise).

Obviously some work is necessary but in the final end - you can easily change the corresponding layer´s Denoising approach later - it is non destructive and re-editable. Topaz is not re-editable in this way. Once you do not like the result you did, you have to go back and reedit denoising from scratch unlike ON1 where you just fine tweak.

I hope I do not repeat the obvious.

It seems I mis-understood your previous message.

Yes I can do what you mentioned with layers, and I am still learning Photo RAW to achieve its best results.