Getting Topaz AI instead of an OM-1 ?

Lez325 wrote:

Original image - LOOK AT BOTH IMAGES AS original SIZE TO SEE THE DIFFERENECES

2 mins in Topaz- a very usable program

Les

Both are  just terrible due to artifacts from AI. I am convinced that RAW processed in Adobe PS with just default color denoise would be significantly better.

Serguei Palto wrote:

Lez325 wrote:

Original image - LOOK AT BOTH IMAGES AS original SIZE TO SEE THE DIFFERENECES

2 mins in Topaz- a very usable program

Les

Both are just terrible due to artifacts from AI. I am convinced that RAW processed in Adobe PS with just default color denoise would be significantly better.

Adobe would be your 1st mistake.

ProDude wrote:

Adobe would be your 1st mistake.

Don't warry on my mistakes. Care on your's mistakes, which seems to be the only your  wealth

Messier Object wrote:

Stefandreas wrote:

X

Just tell me your opinion: to Topaz or not to Topaz ?

My opinion . . .

The OM-1 is not worth the price merely for the minor improvement in high ISO output.

Buy it for the features and speed advantages it may offer you over your E-M1II, not because it has a little better colour fidelity at ISO10,000

I find Topaz Denoise AI to be a useful and convenient tool. It frees me from the tedious task of manually masking my subjects and applying Noise Reduction in Lightroom - which can be tricky with busy background/foregrounds.

But like all tools it requires some practice and skill to get the best results. It has five different AI Denoise ‘models’ each of which can be set to Auto, or you can manipulate the Denoise and sharpening slider controls yourself.

Peter

Um well I guess you missed at least a dozen test and reviews out there on Youtube regarding image tests taken with the OM-1 vs. the previous top of the line in Olympus's lineup. It easily reveals not just lowered noise above ISO1600 but in addition provides better detail without the typical detail destruction that comes with higher ISO's. The color is more accurate as well along with a slight DR boost expected by a BSI/Stacked sensor. It goes without saying processing speeds are greatly upped as well as of course focus lock and tracking. So I wouldn't diminish the advantages of moving up to an OM-1 body. Sure Denoise AI can be an asset as well.

Hi, I tried to treat the Raw files in Photo Lab 6 but am not experienced in denoising. So the results clearly looked worse than the JPEG.

Also I tried to compare treating the Raw file with Topaz AI versus treating the JPEG with Topaz AI - and very very clearly the Topaz worked better on the JPEG than on the Raw. In Raw it washed out all the details whereas in JPEG the results were more subtle.

I really played around with Raw a lot this last days and I cannot confirm your assumption that Raw processing will result in a much better output taking the example below. Maybe a Raw master can but I cannot, spending 30 minutes on Photo Lab 6 newest denoising models. The Raw already contains very funky artefacts which are not easy to correct and frankly: the Olympus JPEG really does a good job. At least better than myself.

Anyway, I think I jump the boat for the OM-1 for the reason of better JPEG in high ISO, overall AF speed and accuracy, animal AF AI, pro capture mode for Leica lenses, better viewfinder. Also EM-10 II as a second body is not good, can’t operate it in the dark, too fiddly, and sensor output is too different from EM-1 II . So first body OM-1 and second EM-1 II makes sense.

Thanks a lot,

Stefan

Serguei Palto wrote:

Lez325 wrote:

Original image - LOOK AT BOTH IMAGES AS original SIZE TO SEE THE DIFFERENECES

2 mins in Topaz- a very usable program

Les

Both are just terrible due to artifacts from AI. I am convinced that RAW processed in Adobe PS with just default color denoise would be significantly better.

ProDude wrote:

Messier Object wrote:

Stefandreas wrote:

X

Just tell me your opinion: to Topaz or not to Topaz ?

My opinion . . .

The OM-1 is not worth the price merely for the minor improvement in high ISO output.

Buy it for the features and speed advantages it may offer you over your E-M1II, not because it has a little better colour fidelity at ISO10,000

I find Topaz Denoise AI to be a useful and convenient tool. It frees me from the tedious task of manually masking my subjects and applying Noise Reduction in Lightroom - which can be tricky with busy background/foregrounds.

But like all tools it requires some practice and skill to get the best results. It has five different AI Denoise ‘models’ each of which can be set to Auto, or you can manipulate the Denoise and sharpening slider controls yourself.

Peter

Um well I guess you missed at least a dozen test and reviews out there on Youtube regarding image tests taken with the OM-1 vs. the previous top of the line in Olympus's lineup. It easily reveals not just lowered noise above ISO1600 but in addition provides better detail without the typical detail destruction that comes with higher ISO's. The color is more accurate as well along with a slight DR boost expected by a BSI/Stacked sensor.

This is not what DPR are saying in thier review of the OM-1. A slight advantage in jpegs but none at all in RAW.

"There's not a big difference in detail levels between the OM-1 and the E-M1 III at low ISOs, though the OM-1's JPEG engine appears to render fine detail a little better.

At higher ISOs the newer camera appears to preserve detail rather better, and gives a rather larger-grained appearance to the noise in areas without detail. Look at the Raw and there's no appreciably difference in noise between the OM-1 and E-M1 III, which is in-line with what we've seen in previous BSI and Stacked CMOS sensors. It's worth noting, though, that the OM-1 keeps up with the older camera's highest ISOs, which is impressive, as this is where we'd expect its more complex design (with four photodiodes per output pixel) to start to count against it."

Trevor Carpenter wrote:

Serguei Palto wrote:

Lez325 wrote:

Original image - LOOK AT BOTH IMAGES AS original SIZE TO SEE THE DIFFERENECES

2 mins in Topaz- a very usable program

Les

Both are just terrible due to artifacts from AI. I am convinced that RAW processed in Adobe PS with just default color denoise would be significantly better.

I'd be very pleased if I had taken that with or without artifacts, great picture

Thank you. That's what I was also thinking. I love this picture and certainly it is not terrible from an aethetic point of you. Print looks gorgeous.

Stefandreas wrote:

Hi, I tried to treat the Raw files in Photo Lab 6 but am not experienced in denoising. So the results clearly looked worse than the JPEG.

Also I tried to compare treating the Raw file with Topaz AI versus treating the JPEG with Topaz AI - and very very clearly the Topaz worked better on the JPEG than on the Raw. In Raw it washed out all the details whereas in JPEG the results were more subtle.

I really played around with Raw a lot this last days and I cannot confirm your assumption that Raw processing will result in a much better output taking the example below. Maybe a Raw master can but I cannot, spending 30 minutes on Photo Lab 6 newest denoising models. The Raw already contains very funky artefacts which are not easy to correct and frankly: the Olympus JPEG really does a good job. At least better than myself.

Anyway, I think I jump the boat for the OM-1 for the reason of better JPEG in high ISO, overall AF speed and accuracy, animal AF AI, pro capture mode for Leica lenses, better viewfinder. Also EM-10 II as a second body is not good, can’t operate it in the dark, too fiddly, and sensor output is too different from EM-1 II . So first body OM-1 and second EM-1 II makes sense.

Thanks a lot,

Stefan

Serguei Palto wrote:

Lez325 wrote:

Original image - LOOK AT BOTH IMAGES AS original SIZE TO SEE THE DIFFERENECES

2 mins in Topaz- a very usable program

Les

Both are just terrible due to artifacts from AI. I am convinced that RAW processed in Adobe PS with just default color denoise would be significantly better.

Let me clarify my point on examples below, which are the hard crops of Raw processed image using different RAW converters. The image is taken at ISO 25600 using G9 camera.

#1. This is the output from Adobe Raw with no denoise. The image does demonstrate an intensive color and luminance noise. At the same time the image shows the maximum details with no artifacts (hair and face skin are realistic)

#2. This is the output from Adobe Raw with the default color denoise. Color noise is completely removed. Some loosing of chrominance is quite visible. All the fine details remain. Artifacts are absent!

#3. This is the output from Adobe Raw with both color denoise (10 units) and luminance denoise (60 units). Due to lower color denoise the chrominance is improved compared to #2. The luminance noise is significantly suppressed and some loosing fine details takes place. Artifacts are absent!

#4. Output from DXO PhotoLab 6 with DeepPrimeXD technology. Color and Luminance noise is well removed. Loosing the chrominance is well visible. Huge number of artifacts, which does not allow to consider the image as acceptable (Hairs are far from realistic; terrible skin artifacts). Actually, the image is ruined!

#5. Output from Topaz DeNoise AI v.3.6.2 with raw model. Color and Luminance noise is well removed. Chrominance is still OK. Huge number of artifacts. Hairs have nothing with the reality, while artifacts differ from the DXO case ; many skin artifacts). The image is killed!

#6. Output from Topaz DeNoise AI v.3.6.2 with low light model. Luminance noise is well removed, but loosing the details in some areas is well visible. Color noise removal is not very good (violet patches remain). Number of artifacts is low, but they still exist at careful study.

Conclusion.

The best in this comparison is Adobe Raw converter - no artifacts; exellent control for the color noise; quite good results with luminance denoise.

Both DXO PL6 and Topaz Denoise AI (raw model) are not acceptable due to artifacts.

The best of AI denoise models is Topaz low light model, which provides lowest number of artifacts. Nevertheless the Adobe RAW is ahead due to the most realistic output.

#1 Hard crop of the image from Adobe PS 2021; no denoise is applyed

#2 Hard crop of the image from Adobe PS 2021; color denoise of 20 units is applyed; No luminance denoise

#3 Hard crop of the image from Adobe PS 2021; color denoise of 10 units is applyed; luminance denoise of 60 units

#4 Hard crop of the image from DXO PhotoLab 6; Denoise by DeepPrimeXD

#5 Hard crop of the image from TopazDenoise AI v.3.6.2; Raw model default +original details restore (50 units)

#6 Hard crop of the image from TopazDenoise AI v.3.6.2; Low light model default +original details restore (50 units)

I am not sure how relevant your observations are to the question posed by the OP, which is a very real one.  I have been trying out the AI software options but don't really feel I can answer that yet.  What is certain is that the software is getting better and better.

I do agree that if looked at at 100% the original image of the buffalo is not very nice.  I am sure, though, that it makes a very satisfactory print at smaller sizes and many people would be happy with it.

Any image is going to fall to pieces if you look at it in sufficient detail, what is important is to compare based on the use case.  I see that you have already cropped the images you use as examples to 0.3Mp (unless you have downrezzed them to that size).  This is way beyond the magnification which would be used by most photographers, but might be relevant in a strictly academic environment.

I see no point in enlarging these images to examine them in more detail.  Based on the images you have presented, my opinion is that #1 and 2 are unusably noisy.  #3 and 6 are too soft.  #4 is the best, and could be quite a usable image, followed by #5 (although #5 does show nasty artifacts if blown up, but for me, that is not a relevant action)

I do agree that we may be approaching this from different viewpoints, I like a photo which looks good even if it you can make valid criticisms of it from an academic point of view.

Each to his/her own!

tom

ahaslett wrote:

tomhongkong wrote:

I am not sure how relevant your observations are to the question posed by the OP, which is a very real one. I have been trying out the AI software options but don't really feel I can answer that yet. What is certain is that the software is getting better and better.

I do agree that if looked at at 100% the original image of the buffalo is not very nice. I am sure, though, that it makes a very satisfactory print at smaller sizes and many people would be happy with it.

Any image is going to fall to pieces if you look at it in sufficient detail, what is important is to compare based on the use case. I see that you have already cropped the images you use as examples to 0.3Mp (unless you have downrezzed them to that size). This is way beyond the magnification which would be used by most photographers, but might be relevant in a strictly academic environment.

I see no point in enlarging these images to examine them in more detail. Based on the images you have presented, my opinion is that #1 and 2 are unusably noisy. #3 and 6 are too soft. #4 is the best, and could be quite a usable image, followed by #5 (although #5 does show nasty artifacts if blown up, but for me, that is not a relevant action)

I do agree that we may be approaching this from different viewpoints, I like a photo which looks good even if it you can make valid criticisms of it from an academic point of view.

Each to his/her own!

tom

I’m not sure that debating which lipstick looks best on a pig is that productive.

Andrew

I agree with that!  It does show though how some of the NR software compare

tom