PL4 DeepPrime versus Topaz DeNoise AI

If you would like to post, you'll need to register. Note that if you have a BCG store account, you'll need a new, separate account here (we keep the two sites separate for security purposes).

BCcanuck

Well-known member
Supporting Member
Marketplace
Both shots processed the same in Photolab 4. No. 1 using DeepPrime No. 2 using DeNoise AI. D500, ISO 4500.
Griz PL4.jpg
You can only see EXIF info for this image if you are logged in.


Griz Topaz DN.jpg
You can only see EXIF info for this image if you are logged in.
 
I'm also a Topaz user but to my eyes there's not much to choose between them. Some areas of #1 look better to me than #2, and vice versa. Personally I'd be satisfied with either :)

Did you set the controls in Topaz and Deep Prime manually or use their auto settings?
 
At first glance I'd say 2 is more appealing but it looks to me like both have a blue/magenta cast that's apparent in the shadow detail like the bear's legs and the rocks.
 
I'm also a Topaz user but to my eyes there's not much to choose between them. Some areas of #1 look better to me than #2, and vice versa. Personally I'd be satisfied with either :)

Did you set the controls in Topaz and Deep Prime manually or use their auto settings?

Auto settings throughout.
 
At first glance I'd say 2 is more appealing but it looks to me like both have a blue/magenta cast that's apparent in the shadow detail like the bear's legs and the rocks.
I think that is a product of the colour of the rocks reflecting in the wet. I remember thinking that they were an unusual colour at the time.
 
I think that is a product of the colour of the rocks reflecting in the wet. I remember thinking that they were an unusual colour at the time.
There's definitely a lot of blue in the rocks as you say, but I'd say the white balance missed a bit and there's an overall blue cast that isn't helping the image.

If you place a color sampler on the black portions of the bear's nose or on the rocks you can see there's an awful lot of blue in those nominally black areas. If you decrease that blue with some color adjustment you'll see that the blacks lose some of that blue cast but the grass color and fur color improve as well.


OriginalColor Temp Adjustment with Camera raw filter
bear-bluecast1.jpg
You can only see EXIF info for this image if you are logged in.
bear-blueremoved.jpg
You can only see EXIF info for this image if you are logged in.

I'm not suggesting the edit above is the 'correct' adjustment, just that there's an overall blue cast that isn't helping the image and may be adding to the sense of unnatural grass color that some folks have mentioned above.
 
Now that you have applied that correction I can see what you mean. I think you are correct about the white balance being "out". It's suprising what different eyes pick up hence the value of forums such as this. Many thanks Dave.
 
I understand that there are differences in processing between the programs - like it can be observed with other software, but what I would be interested in is the PL4 workflow in terms of image filew format. I have done some testing with Topaz Denoise and beside the good result in the final output in terms of noise reduction there were two things that I was irritated about. One was the considerable bending of colors, the other one was breaking up the workflow, because I am forced to change the file format when using Topaz Denoise.

Does PL4 its denoising as part of the non-destructive workflow applied to the NEF file or do you have to convert to DNG or 16bit TIF before doing the denoising ?
Does PL4 leave me with an untouched NEF file with sidecar files allocated to it after post processing including denoising or do I end up with image files in other formats and the door to the processing steps before applying denoising is closed ?

I know there are many out there using PS or LR and they don't (have to) care about this kind of question because DNG is the native Adobe format, but coming from the good old Nikon software using the raw conversion algorithms of the camera manufacturer that many people claim to be superior is a different story.

I have tested with Capture One for Nikon and I was quite enthusiastic after I have read here in the forum that the "for Nikon" version would work with the original raw file algorithms from Nikon. But in reality it happens what always happened in the past with 3rd party software. Part of the information in he NEF file was simply lost. This applies to "simple" things like the active AF point and continues with more complex settings like in-camera image processing parameters that are fully accessible in Nikon software but get lost in the 3rd party products.

I would really like to change over to a software allowing selective processing of images based on masks, areas or alike, but only if I can be sure that I don't loose information in the orginal data and/or or the control for non-destructive processing of the complete information based on the orginal NEF file.

Any comment is highly appreciated and I am looking forward to it as I currently don't have the opportunity for deep diving in POST software.
 
PL4 does noise reduction as part of the demosaicing, so early in the processing workflow. DxO claim that this is the reason for the 'superior' results. PL4 also doesn't touch the NEF file, it writes its data to a sidecar file only. To get a finished product out, you export to JPG or TIFF or any of a number of other formats. However...the Nikon-specific EXIF data like the focus point is not visible in PL4. I don't think it overwrites any of that data - the NEF contains that, after all, and PL4 (along with just about any other sensible software) leaves that file untouched. I don't have Capture NX-D anymore (got tired of the slowness and the bugs) so I cannot check. I'm happy to send you a NEF of an image I've processed with PL4 if you like.

The noise reduction is good, though: this was shot on a Z7 at ISO 12,800...(zoo animals, just practicing). 70-200/2.8 at f/2.8 and 1/500. The full size image is tack sharp.

Jan-Peter

_JPO2129.jpg
You can only see EXIF info for this image if you are logged in.
 
I do not quite understand
" Both shots processed the same in Photolab 4. No. 1 using DeepPrime No. 2 using DeNoise AI..... "
In Photolab you only can use DeepPrime with a RAW file.
Did you prpcess the same RAW file directly in Denoise AI?
It is possible but the results IMH experience ar less than great.
I open Denoise form PS or LR, as .tif files and then I like the results
 
I took a closer look to PL4 after my friend has recommended to do so.
My primary question were three things:
  • Quality of denoising considering that I don't like workflows to be cut up by being forced to change the file format somewhere half way through the way to the final export.
  • Implementation of the U-point technology for selective processing that was originally available in Capture NX2 and eventually got lost due to licensing issues between Nik, Nikon and Google.
  • Being free of keeping my original workflow with directory-based working rather than being forced into things like catalogues, sessions, projects etc. already in the raw processor.
I currently haven't got the time to go into detail (may be I can do that later and topic-related depending on the discussion here), but comparing Capture NX-D, Capture One and PhotoLab4 as programs that are working non-destrucively and directly from native NEF files, I can say that in terms of denoising PL4 is the clear winner (y)
The revival of U-point is great - not only for the old school Capture NX2 fans out there - and from my perspective a breeze to use (y)
Seeing nothing but my drives and directories in the browser after installing the test version was a big relief to me (y):D

Of course the preferences of all of you out there are something very personal, but for people that just want to be able to do the necessary, straight forward processing of photos, PL4 is a good way to go and although I have already spent the money for the Capture One license, I am almost sure I will change over to PL4.
After the programs can access the same NEFs in parallel because they all handle the adjustments in sidecar files, there is nothing lost.

Lets see what's the percentage of pictures processed with which program after a period of time ... ;)
 
FWIIW my workflow. Transfer my RAW NEF file from LR to PL4 via the "transfer" option under file/Plug0in Extras menu option. Denoise in PL4. Return to LR using the export to LR function with noise/optics only. It comes back as a RAW DNG file for further editing in LR. Only the optics/denoise are baked in. If you don't use the optics preset option only the denoise changes are baked in. It is exported to the same folder and (usually) and automatically added to my LR catalog. (There are a few bugs in that process but it usually works fine.) Further editing in LR/ACR or PS or whatever I feel like playing with. Yes the file type changes to DNG, but that does not affect further processing.
 
FWIIW my workflow. Transfer my RAW NEF file from LR to PL4 via the "transfer" option under file/Plug0in Extras menu option. Denoise in PL4. Return to LR using the export to LR function with noise/optics only. It comes back as a RAW DNG file for further editing in LR. Only the optics/denoise are baked in. If you don't use the optics preset option only the denoise changes are baked in. It is exported to the same folder and (usually) and automatically added to my LR catalog. (There are a few bugs in that process but it usually works fine.) Further editing in LR/ACR or PS or whatever I feel like playing with. Yes the file type changes to DNG, but that does not affect further processing.

Aaahh, that's interesting ! So, due to the fact that you use Adobe software you are somebody dealing with DNG anyway, but you are not using Topaz Denoise AI but PL4. It would be interesting to know the reasons. Is it actually the quality of denoising that makes ouy go this way or is it other reasons.

I am not sure whether I did something wrong, but when I tested TopazDenoise AI standalone in combination with Capture NX-D and Capture One I found that it was bending the colours quite significanty in my case. Beside breaking up the workflow this was one of the reasons I decided not to go with Topaz. Butr on the other hand I have seen quite amazing results done with it.
 
I don't normally convert to DNG, that is just how PL4 works. Unless I denoise, my files remain in their native formats. I also have Topaz denoise AI, and have used it in the past, but once I tried (upgraded in my case) to PL4 it became my go to. I don't use PL4 for anything but denoise and it is the only reason I upgraded. I probably will still use Topaz on occasion, but I like my PL4 workflow for those photos I really want to play with.
 
Well Do you want to restore desperately noisy old pictures out from your archives ? You may feel having made a second shooting with a new camera... without spending time and money to return in the same field.

Just try to combinate the following Workflow :

Ingest the original raw into DPLA. Adjust expo/contrast for minimizing the noise on the main subject if possible ( using Upoint is nice. Then apply Deep Prime but with a limited impact on noise reduction such as 30 or 40. Keep little noise for saving most of finest details. Export the tiff into Topaz Denoise and now make a final cleanup low or medium with a bit of sharpness. Then enjoy.

In case of heavy crop and blur : Deep Prime at 40, Tiff into Photoshop ( 1 or 2 steps of separate frequencies enhancement under masking) then final Denoising in Topaz. Finally adjust final sharpness according to print or web size with Steeve method. Some miracles may leave you with a large amount of work for revisiting your old raw archives !
 
I use DXO PL 4 Elite edition to process my Nikon NEF files as it had Deep Prime in it. Essential edition does not. I do not use Topaz. With PL 4 having control points as one set of local adjustments, including sharpening, I no longer use LR. I do admit I am biased towards control points as that is what I used the most with Nikon Capture NX2 when I first went digital. I still use it from time to time as some of its adjustments work better/faster than newer programs.
 
Sorry guys, I know that it is not common practice to quote ones owns posts, but I published some tests around PL4 Deep Prime in another thread accidently, so here they are ...
I just couldn't find a way to move them ....

It is just an example but I think it makes obvious what kind of potential lies withing PL4 with DeepPrime.
Maybe it will provide a little help in relation to software decisions :unsure:

I have used purely Nikon software in the past because of their superior way they interprete the NEF files. Any thrid party software I know up to now ignores parts of the RAW file informartion, like in-camera settings, AF point or such alike. Because I missed the selective processing options that Capture NX2 lost due to the license issue with Nik Software I decided to check out Capture One. Reason was that someone here in the forum said that the Nikon-Only version would work with the native RAW file interpretation from Nikon, which turned out to be wrong. Two things made me finally move away from it and change to DxO PhotoLab 4:
  1. I find the way Capture One tries to "simplify" the workflow and forcing you into their philosophy of catalogues , sessions etc. quite irrititaing for someine like me who is used to organize my workflow and imagae management simply based on directories, naming conventions and keywords in the exported JPG's.
  2. Noise reduction within the program is a bit better than in Capture NX-D, but not the best either. Combining Capture One with tools like Topaz Denoise forces you to break up your workflow, because you have to change file format when entereing Topaz Denoise to TIFF or even DNG. This prevents you from juggling with denoising parameters and other paramters like micro contrast or clarity during processing of picture that require denoising.
In DxO Photolab 4 you have a denoising function completely integrated in the raw processing rather than forsing to use external programs including changing the file fomat and thus make you going away from completely non-destructive processing as soon as you need denoising.

To give you an impression, here are some test picture shot with a D4S and then exported to JPG unprocesses first and then again after applying DeepPrime denoiseing in PhotoLab 4 to it. All 100% crops with no other processing applied than mentioned:

1. ISO 1100 exported as-is and with DeepPrime applied using standard/automatic setting

View attachment 11937View attachment 11938

1. ISO 20.000 exported as-is and with DeepPrime applied using standard/automatic setting

View attachment 11940
View attachment 11941

Remember, all this is not done with external plug-ins, but with the internal denoising function of DxO which allows you to use it in ramdome sequence and combination with the other processing functions. YOu might want to take a look to the next post, where I have another expample (can't attach more than 5 pictures in one post :) )

After also other things in DxO PhotoLab work better for me compared with Capture One, I decided to changeoer to PhotoLab and cancelled the subcriptions of C1 again.
For older images I keep Capture NX-D installed because it features nice support for images that were processed with Capture NX2 using U-Point technology. Capture NX-D is able to show pictures that were processed with Capture NX2 correctly, even if you have used selective processing by means of Control Points / U-Point technology (i.e. selective sharpening of an eye). This is because as opposed to modern programs Capture NX2 was writing the modification IN the raw file itself rather than using sidecar files and Capture NX-D has the capability of interpreting this information correctly.

However, after noise reduction in PhotoLab is so good and the selective processing functions that used to be in Capture NX2 are now implemented in an improved version in PhotoLab 4, I even started to reprocess old images with this software to take advantage of the denoising. Actually DxO bought the technology orginally developped by NIK software back from Google and has implemeted it in their software now.

Beside that I discovered that using Topaz Denoise is bending colours quite a bit, which - in combination with forcing you to change the file format within your workflow makes is relatively difficult to correct that easily.
As promised, here the other - extreme - example with ISO 25.600 plus app 0.7 stops underexpoed as-is / DeepPrime applied with standard settings / DeepPrime applied with standard setting plus reduced micro contrast

View attachment 11944
View attachment 11945

View attachment 11951
 
Hmm.
Usually I use PL4 Denoising first thing on the NEF file and export the resulting file as .dng without any further adjustments back to LR to start the rest of the workflow.
BTW does anybody see a problem with that WF?
I am VERY satisfied with the results. Similar to what you show in your attachments
I noticed also some color bending from Topaz Denoise and also Sharpen. Not always but it did occur.
Thank you for any comments
 
Well, I think it very much depends on what you are used to and in what direction you get guided in the beginning of your post processing "career".

So if you are happy with your workflow, it feels o.k. to work this way and you don't have the impression to loose time by doing it this was and there is nobody complaining about the lack of image quality - due to the procesing, not the shooting itself ;) -, that's all that counts. Said that, my workflow has changed as time went by, because I used different products, sometimes even simultaneously, the necessity of selective processing became bigger, I learned to do this in a different way, ... It is always the question what you really need. The same way it is impossible to name THE best camera of THE best lens, it is impossieble to name THE best post processing software. It always depends what you want to do, what your preferred way of worling is, what the purpose of the images is, ...

In my case the pro nature photographer who brought me to RAW photography was a big fan of the original NEF interpretation and processing algorithms form Nikon and he used View NX2 and Capture NX2 for this reason for a very long time. Adobe's attempt to achieve something similar as with the PDF standard for a RAW image format have lead to what we know as DNG today. As in every sector of software development and data management there is improvement over time and in the early days the loss of information going from any generic RAW format to DNG was pretty considerable. This is logical, because the camera manufacturers know their internal technology best and although any tech company will always be interested in becoming popular and increasing market share by means of supporting established standards, but on the other hand they don't like to be forced into compatibiity to something that isn't a technical standard in the spirit of ISO of IEC standards, but to something another company tries to establish a pseudo-standard basically to earn money by putting pressure on the users to move towards this "standard".

Even today, and without even converting a NEF to a DNG in the first place you can see that the different software packages are interpreting the NEF's differently and this will probably be the same with other manufacturer specific formats. I did some testing around this (Capture NX-D, Capture One, PhotoLab 4). The detailed results will go a bit too far for here and now, but at least most of the non-Nikon software packages loose the information when interpreting NE_F files, being it the information for showing the focus point in post processing or the consequences of in-camera settings for picture control are interpreted differently by different software packages. I know that there are lots of discussions areoud the question which camera setting affects the raw file or not. In my tests it became obvious that settings that have been claimed not to influence the raw data interpretation are indeed having an impact. However, this isn't a big problem as you just have to find out the difference in behaviour between your "old" and your "new" software once and then act accordginly during post processing - as long as the infomation loss does not get critical.

As you saw in the examples it might be worhtwhile juggling with different parameters while trying to find the best compromise between denoising and retaining detail and that is where I see the benefit of PL4 with deep prime for those who prefer working based on the original RAW format rather than going to DNG in the first place.

I have been working with different packages for raw processing now in this sequence:0
  1. View NX2 and Capture NX2
    first, because of the earlier mentioned reasons and the goal not to go for raw on one hand and on the other hand make compromises in quality by moving to another file format in the first place.
  2. View NX-i and Capture NX-D,
    because of the licensing issue between Nikon and Nik software, that made Nikon stop the development of Capture NX2 and thus not provide updates for the program anymore allowing the software to cope with RAW files of more modern camera models. This also made me loose the ability of selective post processing within an image (e.g. brightening, sharpening of a certain area in an image), or at least made me accepting to go away from the homogenous, non-destructive workflow for a relatively small number of pictures.
  3. Capture One,
    because I was looking for a software allowing selective post processing in a homogenous, non-destructive workflow and keeping the information loss for native NEF files to an minimum. This came at the price for a denoising being better that Capture NX-D, but far away from Denoise AI ... or PL4.
  4. PhotoLab 4,
    because of the denoising being superior to everything else I have seen so far - maybe with a slight exception for what @Steve described as his denoising workflow juggling around with masks, layers etc.,
    because the tools for selective processing that I knew from Capture NX2 have come back as part of PL4 with quite some improvements compared to what I knew from the past and
    because the denoising is implemented in a way tat allows random sequence adjustments across all functions within the processing tool box rather than forcing you in a sequence and preventing you from easy readjusting other parameters depending on the result of the chosen denoising method and parameter(s).
After quite some bit of testing I will use PL4 as my primary tool for everything around raw processing, while using XNViewMP for viewing, managing and presenting processed images including management of keywords and IPTC/XMP data. Because I also have to handle tons of business graphics that become integrated in different types of documents and groups of them used for animantions, it was the natural consequence to use the established workflow for their management also for my private photos.
 
PL4 does noise reduction as part of the demosaicing, so early in the processing workflow. DxO claim that this is the reason for the 'superior' results. PL4 also doesn't touch the NEF file, it writes its data to a sidecar file only. To get a finished product out, you export to JPG or TIFF or any of a number of other formats. However...the Nikon-specific EXIF data like the focus point is not visible in PL4. I don't think it overwrites any of that data - the NEF contains that, after all, and PL4 (along with just about any other sensible software) leaves that file untouched. I don't have Capture NX-D anymore (got tired of the slowness and the bugs) so I cannot check. I'm happy to send you a NEF of an image I've processed with PL4 if you like.

Thanks a lot and sorry for the late reaction. In the meantime if have found my way - as you can read above.

Regarding Capture NX-D there are a few findings that might be interesting. A friend of mine, who has been sucessfully working with Capture NX2 for ages now, is now forced to look for an alternative software, because Capture NX2 does not work with the latest Mac OS anymore and also doens not support more modern camera models.
Because he has tons of images processed with Capture NX2 he was scared that he would loose al that work after upgrading his Mac and thus loosig Capture NX2.

So I did some tests for him and the interesting result is that Capture NX-D has one quite unique feature:

Although working non-destructively based on sidecar files it is the only program I know that is able to open NEF files that were processed with Capture NX2 - including selective area processing with the control points and U-Point technology and show them in their correct processed state including the selective adjustments to particular areas in the image.
Because Capture NX-D itself is working on-desctructively, it is also able to apply its complete processing functions to a NEF file that was processed with Capture NX2 before and it correctly applies the settings across the entire image including the areas that have been subject to selective processing.

So, beside all drawbacks it is worthwhile having Capture NX-D on your system, if oyu have a considerable amount of pictures that have been processed with Capture NX2 earlier.

If you take any other of the products I have used, they are able to interpret a NEF file that has been procesed with Capture NX2 before, bit the programs ignore the processing information that Capture NX2 has written into the NEF file. The image will show up in its virgin state as if it came straight from the camera.
 
As you saw in the examples it might be worhtwhile juggling with different parameters while trying to find the best compromise between denoising and retaining detail and that is where I see the benefit of PL4 with deep prime for those who prefer working based on the original RAW format rather than going to DNG in the first place.
Thank you WP for the interesting comments 🙂
Just to clarify, I bring the RAW files from LR to PL4 for denoising, since DeepPrime is only possible on RAW.
I apply Denoise and Optical Corrections via a Preset and export the file back to LR as DNG with only those adjustments via an PL Action.
Then I continue in LR, where I feel more comfortable and, if needed, in PS
 
Back
Top