I have been using Prime (and now Deep Prime) for all my “noisy” images and have been very happy with the results - it seems to me that Deep Prime is currently the best noise reduction function available. The only problem was the very small preview window (which I don’t find useful) and the need to process images and then look at the results. ON1 has just released their NoNoise stand-alone product which (in their usual fashion) promises best in class noise reduction. I was curious - especially since they show the image noise reduction in real time. I bought a copy; the software was released this week and I have been testing it on a couple of my “noisy” images. My conclusion right now is that it is still beta software - very buggy. But, the real time display is very handy for “tuning” the AI noise reduction algorithms. I don’t think DxO has anything to worry about in the short term - but also needs to figure out how to do full image previews of Deep Prime processing.
Another issue with Deep Prime, which is wonderful BTW, is that it only works on RAW Images, and then only on those from cameras It has full optic data on. That leaves out some cameras and I believe all cell phone images. ON1, Topaz and others work on all photos I believe.
ON1 Nonoise will work on Tiff and Jpeg images. However, non raw processing is still a work in progress. When you attempt to use it for that purpose you get a warming that the product is optimized for raw images and that they consider the results with non raw images as beta quality.
Crashed to the desktop twice. Produces output files that are so dark you can’t see anything. This has happened for jpeg, tiff and dng output (it varies - sometimes one, sometimes another. The export menus freeze and the system is not responsive (though it is still using cpu resources). The export menus are confusing and trying to save a preset doesn’t work all of the time (there are multiple “save preset” options but they generally prompt for a new name when you just want to apply updates). The size of the output files are about half the size of the equivalent files output from DxO (for jpegs I have the compression set to 95 for both products). These problems have occurred on both my systems - iMac Pro and MBP16.
In my case it crashed twice while working on different images on Windows 10. It just disappeared and had to be restarted. No other issues and it seemed to work well although in my opinion the results from DeepPRIME were superior. However, as I stated earlier I think NoNoise is faster with much better results when I compared it to my licensed version of Topaz Denoise AI 3. Overall, from my limited testing it seems to be a good product at an appropriate price point.
I tried NoNoise and found it wanting. My results are certainly substantially better than the original file but not close to DeepPRIME. I also found it buggy. The presets for export weren’t set up according to their descriptions and when I actually set the preset to produce a JPEG it produced both a JPEG and a DNG file and the JPEG was really badly off-colour. It looked like a night vision shot, being very dark and entirely green.
They do indicate that the results with non raw files are still in beta. Perhaps they should have waited longer to release this product. I was pleased to find that output DNG files were editable in Photolab. If I recall,correctly DNGs created in Topaz Denoise AI will not open in PhotoLab. It is earlly to draw absolute conclusions but once they have fixed the bugs and finished non raw development I believe this will be a significantly better and faster product than Topaz Denoise AI 3 whichis their main competitor in this market.
I have been testing with raw files (cr2 and arw). I don’t think I would be interested in running it on jpegs. The only thing that keeps me looking is that DxO doesn’t let you preview the (full) images before export. NoNose seems to do a good job of that - works fine on my iMac Pro. But on my (high end) MBP16 it struggles with my A7RM4 61 mPix raw files - the interface freezes up - it is still working but you don’t get any indication that it is still active. You just have to wait.
I agree that On1 isn’t there yet. I just started comparing On1, PureRaw and PhotoLab4. I have found On1 the most inconsistent and generally worse on my files (Canon R5). Badly smearing the details in parts of some images. I like the results on PureRaw for some images but I’m not sure I can live with limitations. For many images the results are over sharpened for my taste. However, I’m having trouble matching the results using PhotoLab4 which makes comparison between these two difficult.
Mark, thanks for responding. I realize, as a newbie to the program names, I misinterpreted the OP. Sorry for the confusion. I thought this was comparing PL4 with PureRaw. In my limited experience the Deep Prime algorithm is far superior to the older Prime (and to On1 NoNoise).
I have been testing PL4 against PureRaw to see if it would be worth paying the additional money for PL4 just for more control over PureRaw’s defaults. My interest in PL4 is having the ability to make changes from the PureRaw defaults for noise, sharpening and amount of lens correction, but it would be much easier to explore making adjustments in PL4 if I had a starting point that matches the PureRaw output.
Matching the output of PhotoLab exactly to PureRAW could be difficult since we don’t know how DXO set the automatic defaults. Besides DeepPRIME (as well as PRIME and HQ), Pureraw also applies chromatic aberration adjustments, lens distortion adjustments, lens sharpness, and vignetting adjustments.
To match against PureRAW, i would start by using the no correction preset in PhotoLab to make sure no edits are applied to your image. Then, use the auto settings (the magic wand) for Chromatic Aberration, Distortion, DXO Denoising, and Vignetting. Use the default settings for Lens Sharpness.
Export the raw image from PhotoLab with these settings using the Export as DNG (Denoise and Optical Corrections only) option and compare the output with the PureRAW DNG version of the image. You may need to adjust the auto settings in Photolab for a closer match. Hope this helps.
Perhaps the thing to realise is this. PureRAW gives you very little control, but whatever it produces you can also produce with PhotoLab.
I would say don’t try to reproduce PureRAW’s results but get PhotoLab to produce what you’re happy with. It can be a little fiddly with the small preview window, but I generally find one or maybe two areas on any photo that will give me a good idea of the result. I am usually impressed with the full results when I finally run the exports. On the odd occasion an area I didn’t consider looks bad so I adjust again with this area now in my preview.
I find it hard to believe in miracles; for the processing to be displayable in real time, the processing speed must be multiplied by fifty.
ON1 would be genius if they succeeded !
And if the process could be made from jpeg pictures, it would be a second miracle ; ON1 specifies that result is better from RAWs (as expected !).
The comparison should be made for me on basis of a complete process ; I would not imagine produce a picture without correcting chromatic aberrations, distortion,… This is a strong point for PL4.
If you want to compare, you could use the same pictures as used in this post.
I think you may have misunderstood my point. I suggested as a starting point for a comparison that he should use the same tools, and only the same tools that PureRAW uses, That includes Denoising, Chromatic Aberration, Distortion, Vignetting, and Lens Sharpness. I also suggested that he set those tools to their default auto values and export the image to DNG using the Export as DNG (Denoise and Optical Corrections only) selection. When you export the file using that selection it only includes the settings for those five tools, assuming they were selected either manually or automatically, No other edits to any other tools are included in that export (except for Volume Deformation if you have a Viewpoint 3 license).That should give a starting point for a comparison to PureRAW’s output.
I too have licensed for fee (with the “early” discount) ON1 NoNoise, and have the Topaz AI noise reduction as well. The ON1 application appears to have numerous software defects and glitches (as noted, and beta). Using both DeepPRIME, ON1, and Topaz Labs current de-noising applications on a MS Win 10 machine, the quality of results for relatively high ISO NEF from a D850, along with detail “restoration”, I find that DP works better on some images, Topaz on others, and ON1 not as well. In terms of output processing wall clock time, in most cases Topaz takes less time (but still slow) compared to DP, and ON1 seems to vary widely (by a factor of three or so). I also tried both of the non-DxO applications on JPEGs (of lower pixel count than the NEFs), DNGs, etc., and these seemed to take less wall clock time than DxO. Note that the times might not be under identical loads depending upon what MS Win was doing in the background – but each was the only applications that I had in use during the comparison tests. As with my past experience with ON1 (before I settled upon PL as my Adobe PS replacement, not willing to use rental software, ON1 does a lot of marketing. ON1 provided a survey before release; after completing the survey, I emailed ON1 a more detaied response I append here: The survey ON1 provided is very difficult to use as it did not have a loupe and depended very much upon the way an individual computer screen is calibrated. I use a (digital) loupe to verify both noise and “sharpness” (by which I mean “detail” – can I see the end of an antenna on a butterfly/moth? Can I see feather structure or eye ring detail in a bird? etc). I made my best guess.
Were the images taken by working photographers using “professional” equipment, or instead designed to look more like what a prosumer might get? End response. (The comment about a prosumer was not meant to be derogatory, but one common attitude is that “one can fix it in post”. For some things this is possible if there is actual “detail” in the pixels; many do not work from raw.) I never received a specific explicit answer. I should note that with DxO support it may take several exchanges before I get a direct answer to a question – but ultimately, I do get a direct answer.
Interesting summary and, at least as it concerns ON1 and DxO, matching with my experience (regarding results).
I was discussing ON1 versus PhotoLab in another forum and another user asked why some of us were so intent on getting the best noise reduction. My response was that I like a clean and sharp look to many of my photos. If there is noise because it was dark, then I expect the darkness to be smooth. But at the same time, DeepPRIME gets rid of ‘invisible’ noise too, resulting in much sharper images. I’ve applied it to low-ISO daylight photos and can see the difference between PRIME and DeepPRIME in the definition of edges and the smoothness of large areas of colour (such as sky).
Before PhotoLab, I used to shoot at lower ISO and ‘accept’ a certain level of noise when needed. So long as the noise was uniform it added ‘character’ to the image. Now I often go back to old photos I originally processed before PhotoLab and reprocess and republish them. I’m in the middle of such a process now. The photos I took while in Singapore in 2019 were originally processed in Luminar 3, then PhotoLab 3 and now PhotoLab 4. Depending on what the team have in store for PhotoLab 5, I may well go back over them yet again. Why? Because the results make me happy.
Thanks for the detailed field report. I have Topaz and Photolab 4. I find the Photolab DeepPrime and previous Prime create very naturalistic results with the settings dialed low. No other noise reduction software manages to tread so lightly, while still cleaning up chroma noise. Luminance noise – as Nikon recognised early – often does not detract from an image. It’s the chroma blots and discolouration (Canon) which ruin an image for mine.
The only reason I can see to use Topaz is that it works on JPEGs, while the different Prime versions only work on RAW. I have used DxO HQ noise reduction on some low light Fuji X-T20 images with good results but the workflow was too tedious:
triage in FastRawViewer
develop Fuji images in Iridient Developer
export to jpg
reprocess in Photolab 4
export to jpg
create and post web versions
It’s only another step and a half really but it was just too much. In the end, it appears noise reduction is best performed simultaneously with the original RAW processing, making Photolab the best and most efficient solution for low light/high ISO photographers. Thanks to all of you for reporting in detail on what’s happening outside our walled garden.