DXO Prime vs DeNoise

Here is another test (in German) of DxO DeepPrime:

Really very impressive and decisive is that the results are available much faster than with Topaz denoise.
And if you get good results with DXO Photolab, you have an excellent all-in-one program.

Again, I think Topaz has a lot of catching up to do, Topaz Studio doesn’t seem to be being developed any more, promised features and above all the seamless integration of the other Topaz programs don’t seem to be a priority anymore. A real pity.
And as already written: the Raw Converter is simply useless, probably not only for Fuji X-Trans. Behind the Raw Converter is probably the LibRaw Dll, at least the result is identical to other programs, which also use the LibRaw (darktable). Here a self-development might be appropriate, now that every year money is flowing into the cashboxes to beat DeepPrime :slight_smile:
Of course I understand that the focus is on the cash cows, like DeNoise, SharpenAI or Video Enhance AI, but I think it would be worthwhile to start the development for other programs again.

2 Likes

I’m trying a trial version of DXO Photolab 4 and have used a very underexposed sony RAW file as a test of Deep Prime noise reduction. In the DXO case I applied phot corrections for the exposure and color sat as well as Deep Prime NR. I also used the same RAW file and developed in Affinity Photo then used the Topaz Denoise AI plugin. I used AI Clear for that. Here are my thought and the photo results below. First, I think DXO should allow a larger preview option for those who have faster machines so that advantage goes to Topaz. DXO does have three NR processes with Deep Prime being the strongest but may not always be necessary. There are some adjustments as well but it is not complicated. For that both are tied.

For the results I brought both jpg files into Affinity photo. I realized that DXO had applied a optical correction so there is a bit of difference in the two. Overall, the results were good for both but I think DXO has a small edge. The carpet is more blurred in Topaz and some light working on the stands almost disappears. See the circled areas on the Topaz results.

The DXO Deep Prime version shows more texture on the carpet and the light wording on the stands is more readable. The wording on the upper sign is bolder on the Topaz version.
I don’t know if it is worth buying DXO Photolab just for the Denoise, even though it is certainly excellent. If you want a pretty good photo program it does have a lot of features that can be worth it. While I did test Denoise AI and the Low Light models in Topaz they didn’t fare better than AI Clear in this case.

Topaz AI clear in Denoise 3:

DXO Deep Prime in Photolab 4:

Here is a screen clip of the undeveloped RAW file to show what I started with: This room was at Biosphere 2 near Tucson, AZ and was too large for my flash plus it had bright light coming through windows.

1 Like

FWIW. I never paid for software to process photos until this past winter. At work, I must use some Adobe products, and am not particularly a fan, but I began to realize that as I moved beyond the capability of my camera to take a great jpg shot, and needed to be able to process the RAW to tweak a little more detail out of an otherwise good shot to make it better (hummingbird wing at high iso/fast shutter speed, or a shot of the Milky Way from southwestern desert skies), then I needed to at least do denoise. I read a lot of reviews last fall, and downloaded 30 day trials of various products, not limited to Topaz Denoise and DXO Photolab 4. By black Friday I had purchased a new laptop with 6gb nVidia graphics (the 10 year old Quattro did not do these softwares any justice), and had settled on Topaz Denoise as a must have, and then also stumbled upon DxO as an option. Both fit my style in different ways. Denoise generally is a very powerful option - but I noticed that when I’d export as DNG, the embedded JPG was tiny and even organizers like FastStone Viewer had trouble. When exporting from RAW to JPG, colors were a bit washed out, too, I thought, but I gradually figured out how to handle that.
As for DxO, I found that I could create a preset for my camera(s), and the lens corrections actually worked well, the horizon correction was good. So I bought the Elite that included the Prime options of Denoise (I did not buy into their Viewpoint add-on, and when the trial expired I missed it - for correcting wonky angles, etc). And as for Topaz, I bought the Black Friday bundle excluding Video Enhance AI (performance was bad, even on a new gaming laptop with 16GB of ram and 6gb GPU) - and I might not use it anyway, so no sense spending the extra money.

My goto for browsing a new day’s folder of images is DXO. I can do a quick cleanup of images, to include denoise if needed, and apply the settings to multiple photos quickly, and I like it for what it does. It’s certainly not the best organizer, may not be the best editor, but it does the basics. However, there are some tough images, that Denoise AI still handles better. And surprisingly, I find GigaPixel very useful for older images from 2-4MP cameras, and it keeps getting better for my purposes. And I like Studio, because it fulfills a niche that neither DXO nor the standalone AI products deliver, which has to do with textures, etc. I don’t use it that often but I do like to use it for certain images.

I guess my point is here that for certain images, one does seem to work better than the other. I like PhotoLab - it fits my style better than many of the other products I trialed, and I can browse, edit, export quickly through several photos. But I also like TL products, including DeNoise, because they can do things PhotoLab cannot do. I spent a lot, a LOT, of time, editing both hummingbird wings and milky way via both products, and in the end, with the images I really wanted to shine, that were on the verge of too blurry, too grainy, or too … something… a combination of Topaz gave me a result I liked.

Here’s one that I processed through TL to get a final result I liked. I used denoise/clear and sharpen/stabilize for this.

Here’s the version I processed through DXO only.

Both are fine, I guess. The colors of DXO look better, but I suppose that’s a reflection on my technique - not quite sure how to get that out of TL. Zoomed in, the detail is better in TL, maybe.

I think many photographers own/use a variety of programs. I like to have a range of “instruments” to use. I gave up on the Photoshop subscription and switched to Affinity Photo which does a very good job and allows plugins from Topaz and my favorite, Luminar AI. I have taken RAW photos for twenty years and find they allow a much greater dynamic range. My workflow is RAW developed in Affinity Photo > Topaz Denoise AI and/or Sharpen AI > Luminar AI > Affinity Photo and save as Jpg.

Your Hummingbird picture is very good.

1 Like

The jury is still out as far as I am concerned. DxO results can be too saturated for my taste, although it does handle colour noise better that DeNoise AI, especially when there are large area that are just one colour.
I am working through some images that I have already processed in DeNoise (pre the latest version), and will offer a couple of real world comparisons later today

Here is an example: Topaz Denoise AI-Severe (auto, colour noise slider pushed as far as practical) on the left; DxO Prime in the middle: original RAW file on the right.
The photo was taken at dawn, ISO was 6,400
DxO seems to be the obvious winner, with far less colour noise on the cobblestones and the purple robes. It has also done some sharpening. I did however have to reduce both saturation and vibrancy, but the greens to my eye are not the correct hue.

When processing a file with moderate noise, I find the results comparable. However, just one example, the original 20MB DNG produced a file of about 80MB in DxO and a whopping 133MB in Topaz DeNoise!

I’d been worried about Photoshop CS6 getting too elderly and started using DXO PhotoLab 4 for several months. The biggest problem is that PL4 doesn’t accept linear DNGs if the metadata indicates the file came from a forbidden camera (Fuji X-Trans). If you use Iridient to convert from the Fuji raw files to DNG, and spoof an acceptable camera, PL4 will accept them, but some features like DeepPrime won’t work. I did find that PL4 works pretty well with consumer-grade, small sensor cameras.

During this same time period I purchased DeNoise AI an once I got a computer with a decent GPU I liked the results. So to compare DeepPrime and DeNoise AI I dug into my files and found some Nikon and Fujifilm (Bayer) RAW and JPEG files and ran them though both. Frankly, for my shooting, I prefer DeNoise AI over DXO DeepPrime. Part of that may be because I started shooting B&W film over 50 years ago when pushing Tri-X to ASA 800 or 1200 was a big deal. Grain doesn’t scare me.

One thing I did discover that the old rule about apply NR as the first step may not hold true with AI NR applications. One of the images I go to is a night sidewalk shot in London’s theatre district. Yesterday I ran the Iridient DNG through PS6 with the intention of using Adjust AI and DeNoise AI. I like the results better when I applied DeNoise AI after I kinda balanced up things with Adjust AI.

Part of my re-look at PS6 is the Nik Collection 4 doesn’t work on any of the legacy PhotoShops – only CC2020/21. So I uninstalled PL4, and did not buy Nik Collection 4. My PS6 plugins are DeNoise AI, Adjust AI, and Nik Collection 3 (primarily for Silver Efex).

I do wish Topaz would bring back the analog film plug-in…But that’s just me.

I switched from PS CS6 (which I used >6 years) to CC when new versions of ACR became non-compatible with old PS, and I really like some features of recent ACR versions. No regrets since that except that I’m staying on PS CC18. My workflow is ACR → PS → Topaz (DAI, SAI, and some older ones) in plug-in mode on a layer copy with further application to background via a layer mask filtering. And I use Nik Collection 3 in the same way. Like you, I found PL4&DeepPrime results inferior for my purposes.