Noise Removal - total time

[Jestertheclown]

I think most people who have LrC from now on will not bother with DXO or Topaz for processing raw files. Most won't care about whether for that for a particular photo one may do slightly better, but for another photo not. LrC Denoise AI is built-in and included for no additional cost. Seems like PureRaw really has no reason to exist anymore.

Yes, that's probably true for LRC users, but perhaps PureRAW still has a role with C1 users?

I doubt it.

In fact, I could never see the point of PureRaw. It was only ever a very expensive one-trick-pony.

I suspect many Capture One users will also be users of Photoshop so they'll have the Adobe noise AI stuff in there.

But it was probably always destined to be a short life product, as it targetted a very specific, and probably temporary, gap in a couple of competitor products. DxO must have known that Adobe and C1 would fill that very well known gap sooner or later.

Topaz can also handle non-raw files which Denoise AI and PureRaw cannot so that still is useful. Adobe says they are working on making Denoise AI work with non-raw files though.

For people who do not use LrC then DXO and Topaz can still sell to them.

As a PhotoLab user, I really hope DxO returns to focusing more on it, and less on PureRAW (I felt the same about another short-lived product, DxO ONE).

And Topaz seems to have stopped working on DeNoise AI some time ago, as it probably realised that its life was finite. Now, it's focused entirely on Photo AI, which is still saleable to users of products with good NR. It still has the lead on AI sharpening and upsizing, but of course that lead may also vanish one day.

Photo AI is turning out to be quite a remarkable bit of gear.

I agree that although I rarely use it, its sharpening works very well and I only occasionally upsize anything so I can't comment but its noise reduction is up there with the best.

Obviously, your love affair with DxO will blinker you to that but in real life, Topaz does an excellent job.


"It's good to be . . . . . . . . . Me!"

 

[Digital Nigel]

I think most people who have LrC from now on will not bother with DXO or Topaz for processing raw files. Most won't care about whether for that for a particular photo one may do slightly better, but for another photo not. LrC Denoise AI is built-in and included for no additional cost. Seems like PureRaw really has no reason to exist anymore.

Yes, that's probably true for LRC users, but perhaps PureRAW still has a role with C1 users?

I doubt it.

In fact, I could never see the point of PureRaw. It was only ever a very expensive one-trick-pony.

I suspect many Capture One users will also be users of Photoshop so they'll have the Adobe noise AI stuff in there.

But it was probably always destined to be a short life product, as it targetted a very specific, and probably temporary, gap in a couple of competitor products. DxO must have known that Adobe and C1 would fill that very well known gap sooner or later.

Topaz can also handle non-raw files which Denoise AI and PureRaw cannot so that still is useful. Adobe says they are working on making Denoise AI work with non-raw files though.

For people who do not use LrC then DXO and Topaz can still sell to them.

As a PhotoLab user, I really hope DxO returns to focusing more on it, and less on PureRAW (I felt the same about another short-lived product, DxO ONE).

And Topaz seems to have stopped working on DeNoise AI some time ago, as it probably realised that its life was finite. Now, it's focused entirely on Photo AI, which is still saleable to users of products with good NR. It still has the lead on AI sharpening and upsizing, but of course that lead may also vanish one day.

Photo AI is turning out to be quite a remarkable bit of gear.

I agree that although I rarely use it, its sharpening works very well

I've gone back to using Sharpen AI, which still does a better job than Photo AI. I'm quite surprised that Photo AI's sharpening hasn't overtaken the older product yet. It actually went backwards with 1.3.0, though 1.3.1 partially fixes it.

I like the idea of simplifying the UI in Photo AI, but in practice, they seem to get it wrong much too often.

and I only occasionally upsize anything

Same here, but Photo AI seems to have overtaken Gigapixel now. It includes facial reconstruction which is a bit hit or miss, but sometimes useful with poor quality scanned photos.

so I can't comment but its noise reduction is up there with the best.

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

Obviously, your love affair with DxO will blinker you to that but in real life, Topaz does an excellent job.

I have the latest versions of both, and use whichever is best. I also have the latest version of Luminar Neo, but that's seldom the best at anything, so I usually forget to even try it.

 

[Ephemeris]

I think most people who have LrC from now on will not bother with DXO or Topaz for processing raw files. Most won't care about whether that for a particular photo one may do slightly better, but for another photo not. LrC Denoise AI is built-in and included for no additional cost. Seems like PureRaw really has no reason to exist anymore. Topaz Denoise AI can also handle non-raw files which Denoise AI and PureRaw cannot so that still is useful. Adobe says they are working on making Denoise AI work with non-raw files though.

For people who do not use LrC then DXO and Topaz can still sell to them.

Speed give Topaz a big edge (it's factors faster looking across our own and results here) and also it will auto upscale and resolve focus issues. Our view is if we had a pc suitable for Adobe Denoise (don't know what's best) it's still slower than Adobe so batch processing maybe impacted (but maybe it's not a problem)

My post was not about whether one program is faster or produces slightly better results for some photos. It was about this:

I think most people who have LrC from now on will not bother with DXO or Topaz for processing raw files.

So for those that need other features and/or don't enjoy the speed then surely they will bother with other programmes.

I do think that having it built-in at no additional cost and also much better integrated into LrC will result in most people not bothering with add on products from DXO and Topaz. Note that I used the word most, not the word all.

 

[sybersitizen]

Topaz AI still does NR?

Are you asking about Topaz Photo AI? Yes, of course it still does. That's one of its reasons for being.

It's slightly slower than DeNoise AI which was last updated in Feb.

if it's slower, that's presumably because it does more things.

 

[Ephemeris]

Topaz AI still does NR?

Are you asking about Topaz Photo AI? Yes, of course it still does. That's one of its reasons for being.

It's a rhetorical question where it had seemed to be suggested that Topaz are not updating Noise tools now.

It's slightly slower than DeNoise AI which was last updated in Feb.

if it's slower, that's presumably because it does more things.

You can have only NR switched on. But some recent tests by members have shown it to be a little slower than DeNoise AI which is an interesting find. For those of us that batch process aot of images its perhaps helpful to know.

You are likely very correct in that it is doing more, but more what I am not sure.

 

[Jestertheclown]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

Obviously, your love affair with DxO will blinker you to that but in real life, Topaz does an excellent job.

"It's good to be . . . . . . . . . Me!"

 

[Horacecoker]

How about now? Amount 60. All edits have been removed from original. Also added contrast and some texture on the lion.

Yes, that's definitely an improvement overall, though it's noisier. But compared to PL6, it's much noisier, with the noise obscuring some of the visible detail and softer.

It strikes me that it's too much to expect a first generation product like LR DN AI to compete with a much more mature, second generation product like DeepPRIME XD. So, I thought it might be fairer to compare with the technology it's a copy of, the older DeepPRIME:


PL DeepPRIME. This tech was introduced in 2020 in PL4

Comparing them:

And, guess what? Even DxO's older first generation AI NR produces an image with less noise, more details and more sharpness. And it's quicker, too. So Adobe hasn't even caught up with PL4.

Why does this comparison which you say is with DeepPrime, state on the actual crop it's PL DP XD?

Ah, thanks, yes, it looks I inserted the wrong comparison shot. Try this (which is the one my comments were based on):


LR60 vs DP80

And it's totally farcical to suggest DeepPrime (normal or XD) is far superior or even just better than LR AI Denoise with this single terribly off-colour, magenta-soaked image taken at a stupidly high ISO.

I've yet to see an image where DeepPRIME XD didn't do a better job than the new LR AI NR (less noise, more detail), but I agree that this very high ISO shot is a particularly tough test of a primitive first gen product like LR AI NR. PL6, however, had no problems with it, even using the older DeepPRIME technology from PL4. But I did raise the NR setting more than I would with my normal high ISO images (I'd never go so high).

Of course, this isn't really a good comparison images as, because of the high depth of field, much of it is out of focus, so it's hard to compare the rendering of details; a flat, in-focus target would be better Also, being only a large animal suit meant to be viewed from a distance, there's no natural fur detail to reproduce.

Here's a crop of an actual image of mine, taken from a moving underground airport shuttle, hence the high shutter speed. ISO16000 is as high as I would choose to use, and it's processed using my default XD settings:


PL6, DeepPRIME XD 50

[Yes, those are rain water droplets on the windscreen, as it was drizzling outside]

The raw image:
https://www.dropbox.com/s/4p81p6woy9q6154/Airport transit A6402283.ARW?dl=0

What can LR do with this image?

I'll have a go at that image shortly. In the meantime here's the magenta mess image with LR DN at 80%. Equally as good as 80% DeepPrime for my money.


LR Denoise at 80%

That's certainly better than 60, but it's still slightly noisier, clearly less detailed and softer than DeepPRIME. You've obviously lost detail as the price for reducing the noise to an acceptable level (a compromise not needed with DeepPRIME). Incidentally, I didn't do any sharpening in PL, apart from DxO's automatic distortion correction lens sharpening, but it's still a sharper image.

I'd have thought your version would probably be acceptable to most people (particularly considering the very high ISO), apart from perfectionists…

I'm pretty much a perfectionist and it's for that reason I would never shoot at these silly ISOs.

Here's my attempt with your underground image I settled for 85% in LR Denoise.


LR DN 85%

Thanks, I'll compare them tomorrow when I'm on my big screen computer (my 11" iPad Pro screen is too small).

I won't be doing anymore. They take nearly 6 minutes on my 7-year-old PC with a paltry 2gb GPU. I need a new computer!

Wow, that is slow indeed! My PC dates from late 2018, and back then, I didn't think to specify a top-end GPU, as I didn't have any products back then that made much use of the GPU (mine has 6GB).

Next time, I'll think much more about the GPU, and less about the CPU. It was the latter that mattered most with the much older (non-AI) PRIME technology.

My old machine takes about 12-15 secs to process a DeepPRIME 20-24mp image, and about twice that for XD. But that's effectively zero waiting time, as the processing happens in the background, and doesn't stop me working on the next image.

I guess the latest GPUs might be perhaps five times faster than my old 1060 GPU, but as the current speed doesn't get in my way, I can live with the current performance.

As I'm due a new PC anyway, I've done a lot of research and I'm now in the process of speccing out a custom build on Scan Computers website. (They are based only a few miles from me). Top-of-the-range Nivida GPUs are silly money and only really 'necessary' for ardent 4k gamers. The two mid-range GPUs I'm looking at are

12GB NVIDIA GeForce RTX 3060 or 8GB NVIDIA GeForce RTX 3060 Ti

From what I've read these two come highly recommended for Photoshop etc.

 

[Digital Nigel]

How about now? Amount 60. All edits have been removed from original. Also added contrast and some texture on the lion.

Yes, that's definitely an improvement overall, though it's noisier. But compared to PL6, it's much noisier, with the noise obscuring some of the visible detail and softer.

It strikes me that it's too much to expect a first generation product like LR DN AI to compete with a much more mature, second generation product like DeepPRIME XD. So, I thought it might be fairer to compare with the technology it's a copy of, the older DeepPRIME:


PL DeepPRIME. This tech was introduced in 2020 in PL4

Comparing them:

And, guess what? Even DxO's older first generation AI NR produces an image with less noise, more details and more sharpness. And it's quicker, too. So Adobe hasn't even caught up with PL4.

Why does this comparison which you say is with DeepPrime, state on the actual crop it's PL DP XD?

Ah, thanks, yes, it looks I inserted the wrong comparison shot. Try this (which is the one my comments were based on):


LR60 vs DP80

And it's totally farcical to suggest DeepPrime (normal or XD) is far superior or even just better than LR AI Denoise with this single terribly off-colour, magenta-soaked image taken at a stupidly high ISO.

I've yet to see an image where DeepPRIME XD didn't do a better job than the new LR AI NR (less noise, more detail), but I agree that this very high ISO shot is a particularly tough test of a primitive first gen product like LR AI NR. PL6, however, had no problems with it, even using the older DeepPRIME technology from PL4. But I did raise the NR setting more than I would with my normal high ISO images (I'd never go so high).

Of course, this isn't really a good comparison images as, because of the high depth of field, much of it is out of focus, so it's hard to compare the rendering of details; a flat, in-focus target would be better Also, being only a large animal suit meant to be viewed from a distance, there's no natural fur detail to reproduce.

Here's a crop of an actual image of mine, taken from a moving underground airport shuttle, hence the high shutter speed. ISO16000 is as high as I would choose to use, and it's processed using my default XD settings:


PL6, DeepPRIME XD 50

[Yes, those are rain water droplets on the windscreen, as it was drizzling outside]

The raw image:
https://www.dropbox.com/s/4p81p6woy9q6154/Airport transit A6402283.ARW?dl=0

What can LR do with this image?

I'll have a go at that image shortly. In the meantime here's the magenta mess image with LR DN at 80%. Equally as good as 80% DeepPrime for my money.


LR Denoise at 80%

That's certainly better than 60, but it's still slightly noisier, clearly less detailed and softer than DeepPRIME. You've obviously lost detail as the price for reducing the noise to an acceptable level (a compromise not needed with DeepPRIME). Incidentally, I didn't do any sharpening in PL, apart from DxO's automatic distortion correction lens sharpening, but it's still a sharper image.

I'd have thought your version would probably be acceptable to most people (particularly considering the very high ISO), apart from perfectionists…

I'm pretty much a perfectionist and it's for that reason I would never shoot at these silly ISOs.

Here's my attempt with your underground image I settled for 85% in LR Denoise.


LR DN 85%

Thanks, I'll compare them tomorrow when I'm on my big screen computer (my 11" iPad Pro screen is too small).

I won't be doing anymore. They take nearly 6 minutes on my 7-year-old PC with a paltry 2gb GPU. I need a new computer!

Wow, that is slow indeed! My PC dates from late 2018, and back then, I didn't think to specify a top-end GPU, as I didn't have any products back then that made much use of the GPU (mine has 6GB).

Next time, I'll think much more about the GPU, and less about the CPU. It was the latter that mattered most with the much older (non-AI) PRIME technology.

My old machine takes about 12-15 secs to process a DeepPRIME 20-24mp image, and about twice that for XD. But that's effectively zero waiting time, as the processing happens in the background, and doesn't stop me working on the next image.

I guess the latest GPUs might be perhaps five times faster than my old 1060 GPU, but as the current speed doesn't get in my way, I can live with the current performance.

As I'm due a new PC anyway, I've done a lot of research and I'm now in the process of speccing out a custom build on Scan Computers website. (They are based only a few miles from me). Top-of-the-range Nivida GPUs are silly money and only really 'necessary' for ardent 4k gamers. The two mid-range GPUs I'm looking at are

12GB NVIDIA GeForce RTX 3060 or 8GB NVIDIA GeForce RTX 3060 Ti

From what I've read these two come highly recommended for Photoshop etc.

Yes, from what I've read, those sound like very good choices. They're both a lot faster than my old GTX1060 6GB. I think the key is to get an RTX card with the Tensor cores.

Like you, I have my PCs custom built by a local firm (Armari) that specialises in high performance workstations. I might ask their advice about a sensible upgrade for my GPU (the rest of the machine will be good for years to come). I know the power supply can be limiting factor.

 

[Digital Nigel]

I've yet to see an image where DeepPRIME XD didn't do a better job than the new LR AI NR (less noise, more detail), but I agree that this very high ISO shot is a particularly tough test of a primitive first gen product like LR AI NR. PL6, however, had no problems with it, even using the older DeepPRIME technology from PL4. But I did raise the NR setting more than I would with my normal high ISO images (I'd never go so high).

Of course, this isn't really a good comparison images as, because of the high depth of field, much of it is out of focus, so it's hard to compare the rendering of details; a flat, in-focus target would be better Also, being only a large animal suit meant to be viewed from a distance, there's no natural fur detail to reproduce.

Here's a crop of an actual image of mine, taken from a moving underground airport shuttle, hence the high shutter speed. ISO16000 is as high as I would choose to use, and it's processed using my default XD settings:


PL6, DeepPRIME XD 50

[Yes, those are rain water droplets on the windscreen, as it was drizzling outside]

The raw image:
https://www.dropbox.com/s/4p81p6woy9q6154/Airport transit A6402283.ARW?dl=0

What can LR do with this image?

I'll have a go at that image shortly. In the meantime here's the magenta mess image with LR DN at 80%. Equally as good as 80% DeepPrime for my money.


LR Denoise at 80%

That's certainly better than 60, but it's still slightly noisier, clearly less detailed and softer than DeepPRIME. You've obviously lost detail as the price for reducing the noise to an acceptable level (a compromise not needed with DeepPRIME). Incidentally, I didn't do any sharpening in PL, apart from DxO's automatic distortion correction lens sharpening, but it's still a sharper image.

I'd have thought your version would probably be acceptable to most people (particularly considering the very high ISO), apart from perfectionists…

I'm pretty much a perfectionist and it's for that reason I would never shoot at these silly ISOs.

Here's my attempt with your underground image I settled for 85% in LR Denoise.


LR DN 85%

OK, I've now had a chance to take a look at that. I see it has less noise than I got with XD at 50, so I re-ran it with XD at 60, which produced about the same (completely acceptable) noise level as your image. Your image is a little brighter than mine (both are brighter than real life). Comparing a crop:




LR 85 compared with DPXD at 60

There isn't a huge difference, but I think the XD version is a little sharper and more detailed. I didn't apply any sharpening, other than the usual DxO lens correction.

I won't be doing anymore. They take nearly 6 minutes on my 7-year-old PC with a paltry 2gb GPU. I need a new computer!

My rendering took 50 seconds on my 4.5 year-old computer with its GTX1060 with 6GB. I estimate it would be under 10 seconds with a modern RTX GPU.

 

[Digital Nigel]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

OK, here you go:

Loading…

www.dropbox.com

What's the best you can achieve with this image in Topaz and LR? This is what I produced in PL6 (XD, increased to 70) and Photo AI (increased to 70 from the much lower suggested level):




PL 6.5, DP XD at 70






Topaz Photo AI 1.3.1, strength increased to 70; only NR applied



Can you beat the PL6 result in any other product of your choice?

Incidentally, it was this image that persuaded me to upgrade from PL5 to PL6. I was, and remain, very happy with DeepPRIME as found in PL4/5, and wasn't convinced that XD was much of an improvement (and there weren't many other enhancements that attracted me to PL6). This image proved to me that XD was needed for some images.

 

[Jestertheclown]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

OK, here you go:

https://www.dropbox.com/s/47m2wx664dzlqya/Frigate in Rhodes R1000819.ARW?dl=0

What's the best you can achieve with this image in Topaz and LR? This is what I produced in PL6 (XD, increased to 70) and Photo AI (increased to 70 from the much lower suggested level):


PL 6.5, DP XD at 70


Topaz Photo AI 1.3.1, strength increased to 70; only NR applied

Can you beat the PL6 result in any other product of your choice?

Incidentally, it was this image that persuaded me to upgrade from PL5 to PL6. I was, and remain, very happy with DeepPRIME as found in PL4/5, and wasn't convinced that XD was much of an improvement (and there weren't many other enhancements that attracted me to PL6). This image proved to me that XD was needed for some images.

I've got far better things to do than become involved in your juvenile games.

Having declined your challenge, I'll await your other default setting; name calling when you don't get your own way.


"It's good to be . . . . . . . . . Me!"

 

[Digital Nigel]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

OK, here you go:

https://www.dropbox.com/s/47m2wx664dzlqya/Frigate in Rhodes R1000819.ARW?dl=0

What's the best you can achieve with this image in Topaz and LR? This is what I produced in PL6 (XD, increased to 70) and Photo AI (increased to 70 from the much lower suggested level):


PL 6.5, DP XD at 70


Topaz Photo AI 1.3.1, strength increased to 70; only NR applied

Can you beat the PL6 result in any other product of your choice?

Incidentally, it was this image that persuaded me to upgrade from PL5 to PL6. I was, and remain, very happy with DeepPRIME as found in PL4/5, and wasn't convinced that XD was much of an improvement (and there weren't many other enhancements that attracted me to PL6). This image proved to me that XD was needed for some images.

I've got far better things to do than become involved in your juvenile games.

Having declined your challenge, I'll await your other default setting; name calling when you don't get your own way.

Hopefully someone else has more confidence in their preferred products and skills?

 

[sybersitizen]

Topaz AI still does NR?

Are you asking about Topaz Photo AI? Yes, of course it still does. That's one of its reasons for being.

It's a rhetorical question where it had seemed to be suggested that Topaz are not updating Noise tools now.

It's slightly slower than DeNoise AI which was last updated in Feb.

if it's slower, that's presumably because it does more things.

You can have only NR switched on.

Even with only NR activated, it does sharpening. In fact, that sharpening is often enough, so my default is to leave the Sharpening panel disabled.

But some recent tests by members have shown it to be a little slower than DeNoise AI which is an interesting find. For those of us that batch process aot of images its perhaps helpful to know.

You are likely very correct in that it is doing more, but more what I am not sure.

If DeNoise AI (which I've never used) is not doing the same 'hidden' sharpening, that could be one thing.

 

[Digital Nigel]

Topaz AI still does NR?

Are you asking about Topaz Photo AI? Yes, of course it still does. That's one of its reasons for being.

It's a rhetorical question where it had seemed to be suggested that Topaz are not updating Noise tools now.

It's slightly slower than DeNoise AI which was last updated in Feb.

if it's slower, that's presumably because it does more things.

You can have only NR switched on.

Even with only NR activated, it does sharpening. In fact, that sharpening is often enough, so my default is to leave the Sharpening panel disabled.

But some recent tests by members have shown it to be a little slower than DeNoise AI which is an interesting find. For those of us that batch process aot of images its perhaps helpful to know.

You are likely very correct in that it is doing more, but more what I am not sure.

If DeNoise AI (which I've never used) is not doing the same 'hidden' sharpening, that could be one thing.

DeNoise AI does do some sharpening, just as Sharpen AI does some NR. It's because these two processes interact (just removing noise tends to soften images, so there needs to be some compensatory sharpening, while sharpening also accentuates the noise, which needs to be reduced first). That was one reason to combine both in one tool.

 

[Ephemeris]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

OK, here you go:

https://www.dropbox.com/s/47m2wx664dzlqya/Frigate in Rhodes R1000819.ARW?dl=0

What's the best you can achieve with this image in Topaz and LR? This is what I produced in PL6 (XD, increased to 70) and Photo AI (increased to 70 from the much lower suggested level):


PL 6.5, DP XD at 70


Topaz Photo AI 1.3.1, strength increased to 70; only NR applied

Can you beat the PL6 result in any other product of your choice?

Incidentally, it was this image that persuaded me to upgrade from PL5 to PL6. I was, and remain, very happy with DeepPRIME as found in PL4/5, and wasn't convinced that XD was much of an improvement (and there weren't many other enhancements that attracted me to PL6). This image proved to me that XD was needed for some images.

I've got far better things to do than become involved in your juvenile games.

Having declined your challenge, I'll await your other default setting; name calling when you don't get your own way.

Hopefully someone else has more confidence in their preferred products and skills?

As a guess that's not the reason.

I had thought it may help to include the original for comparison.

To have an improved result would need some measurable requirements rather than preference. What are you trying to achieve?

 

[Serguei Palto]

It's certainly not as good as PL with raws, but is probably the best for JPEGs.

That's not been my experience and stating the above as if it was a fact doesn't help anyone.

As I've said (many times) your fanatical devotion to DxO makes you so biased that your default mode is to disagree with anyone who dares to suggest that it can be bettered.

OK, here you go:

https://www.dropbox.com/s/47m2wx664dzlqya/Frigate in Rhodes R1000819.ARW?dl=0

What's the best you can achieve with this image in Topaz and LR? This is what I produced in PL6 (XD, increased to 70) and Photo AI (increased to 70 from the much lower suggested level):


PL 6.5, DP XD at 70


Topaz Photo AI 1.3.1, strength increased to 70; only NR applied

Can you beat the PL6 result in any other product of your choice?

Incidentally, it was this image that persuaded me to upgrade from PL5 to PL6. I was, and remain, very happy with DeepPRIME as found in PL4/5, and wasn't convinced that XD was much of an improvement (and there weren't many other enhancements that attracted me to PL6). This image proved to me that XD was needed for some images.

Any standard processing just with color noise reduction will result in much better result than yours example DXO processing. The reason is simple. Your image contains huge number of artifacts which, for some reason, are not seen by your. For example, look at this crop from your image.




Crop from your DXO-processed image



BTW, I have nothing against DXO. Moreover, in particular case I could recommend you to switch-off the lens sharpening, so you can get quite decent result from DXO DeepPrime, where artifacts are not so visible (see next crop).




DXO DeepPrime AI with no lens sharpening (PL-6)

A result from non-AI processing looks as follows, where no artifacts indeed. Only photos with no artifacts have their value (cost).




non-AI noise reduction and no sharpening

The result from RAW with no noise reduction is as follows. Color noise is high, but this image has no artifacts, so it has the cost.




No noise reduction

 

[Sagittarius]

The result from RAW with no noise reduction is as follows. Color noise is high, but this image has no artifacts, so it has the cost.


No noise reduction

What cost does it have? You cannot see anything. Also isn't this color noise artifact itself?

 

[Ephemeris]

The result from RAW with no noise reduction is as follows. Color noise is high, but this image has no artifacts, so it has the cost.


No noise reduction

What cost does it have? You cannot see anything. Also isn't this color noise artifact itself?

It's not an artifact from the tool I think is the point.

Is it an innacuracy, and one that can be categorised then yes (I think this is what you meant anyway).

 

[Ephemeris]

Topaz AI still does NR?

Are you asking about Topaz Photo AI? Yes, of course it still does. That's one of its reasons for being.

It's a rhetorical question where it had seemed to be suggested that Topaz are not updating Noise tools now.

It's slightly slower than DeNoise AI which was last updated in Feb.

if it's slower, that's presumably because it does more things.

You can have only NR switched on.

Even with only NR activated, it does sharpening. In fact, that sharpening is often enough, so my default is to leave the Sharpening panel disabled.

You can switch off NR

You can switch off sharpening

If you do both you will see that the previews are identical and it doesn't do any processing.

Take a small crop of an image with some fine hair for example. Turn both options off. Save the output.

Turn sharpening on at 1. Save. Increase by 1. Save. Stop at 10.

Compare those 10 images with the original. 2 will look the same.

But some recent tests by members have shown it to be a little slower than DeNoise AI which is an interesting find. For those of us that batch process aot of images its perhaps helpful to know.

You are likely very correct in that it is doing more, but more what I am not sure.

If DeNoise AI (which I've never used) is not doing the same 'hidden' sharpening, that could be one thing.

 

[Serguei Palto]

The result from RAW with no noise reduction is as follows. Color noise is high, but this image has no artifacts, so it has the cost.


No noise reduction

What cost does it have? You cannot see anything. Also isn't this color noise artifact itself?

"You can not see anything..." All depends on who looks. For example, in science the usefull information is very often extrated from extremely noisy signal.

We deal with PHOTO GRAPHY (photo=light). Noise comes from nature of light. You can not escape the noise as soon as you deal with light. The true denoise methods can clarify (make more accurate) the information by removing the noise in some spectral range where the information is absent. But if someone has artificially introduced NEW INFORMATION, then we no more deal with the photography. After artifacts are introduced you can not more extract the true information which came from light. The image losts its cost.

Hope, this will help.