I've owned lots of image de-noisers over the years. I've used external apps, Photoshop/Lightroom plugins, and I've used the noise reductions controls built into Photoshop and Lightroom. They all work to some degree, but you wind up trading noise reduction for details, and I've never found a noise reduction aid that didn't have some rather large trade-offs. Increase the noise reduction and details suffer. Preserve the details and the noise reduction is not effective. Many of you know the drill.
Over the years, I used the Topaz DeNoise plugin with fair to good results, and now, it's been updated to Topaz DeNoise AI. Its claim to fame is Topaz inspected thousands of noisy images and added AI smarts to help this new app/plugin know the difference between noise and actual data in an image.
Most of my imaging is in low light, as I do almost all my landscape work around sunrise and sunset. Working in reduced light and at high ISOs are an invitation to noise. Camera sensors are getting better and better, but there was not a single image I fed DeNoise AI that was not improved, even when inspecting the image at high magnifications.
I also do a lot of night photography, capturing the Milky Way, and I also do some telescope-based imaging. Looking at those images, again, I found DeNoise AI significantly reduced the noise with no visible degradation of detail. You can see below how the sky is cleaned up in this enlargement without doing damage to the stars. (It may be hard to see the improvement in these samples because I'm limited to small compressed files for uploading, but you can download large samples at the Topaz site.)
I can report that if you use the sliders in DeNoise AI aggressively, you're going to lose some detail, add some halos, or increase sharpness to the degree that you may see some micro-image breakup. Having said that, I never had to push those sliders very hard, because using the split screen feature and looking at the image at 400 percent magnification, I could easily find a setting where image grain (noise) was made invisible and I was not losing detail.
Topaz offers a 30-day free trial of the software, so you can judge for yourself. Until May 3rd, the software is $59.00, and after that, it jumps back to a list price of $79.00. Users of the older Topaz DeNoise are getting offers of a free upgrade, so check the Topaz site and see if you qualify or watch for an email offer.
I found Topaz DeNoise AI a significant achievement in noise reduction. It's not completely automatic, and the processing is not instant; it can take several seconds to see the result, but the wait is worth it,
If you are a photographer fighting noise, I suggest you try the free trial of DeNoise AI. No noise reduction tool is 100 percent effective, but this one comes really, really close. I'm not easily impressed, and this noise solution impressed me.
Have you tried other apps like Noise Ninja or Imagenomic Noiseware? Wasn't sure if this article was a paid product placement ad since there's no mention of similar alternatives. (Both are very good as well)
I just tried DeNoise AI on some 12800 ISO images I shot at the theater, and the noise reduction is much better than the latest Camera Raw, but previews and saving, as Paul stated, were definitely slow. Another downside is the DNG file I saved was 100Mb larger than the original CR2 file. It would also be nice if there was a batch process feature, so you could let your system churn through a lot of similar photos.
"I've never found a noise reduction aid that didn't have some rather large trade-offs"
It would be helpful to know what you have used. I suspect you haven't used DxO PhotoLab 2 Elite, widely regarded as having perhaps the best noise reduction among RAW processors available today. I tested PhotoLab's predecessor (Optics Pro), Aperture, Capture One and Lightroom several years ago, and DxO outclassed the rest in noise reduction by a large margin. I was using Topaz DeNoise at the time and found it very effective, but it required a great deal of user intervention for every single image, whereas DxO produced equally good or better results with no user intervention at all.
Very interesting article!
I just published a test a made about the best noise reduction software (Testing Loghtroom, Photoshop, Luminar, DXO, Capture One, Dfine 2, Noiseware, Topaz Denoise, Noise Ninja and Neat image pro) and the best at eliminating noise in skies while preserving details was without a doubt Noiseware. Below you can see the graphic results after testing the same image with all types of software: -reduction-software/
Topaz may be their own worst enemy in that several of their products have denoise sliders. And "Clairity AI" is made specifically for noise (and sharpness), leaving a person to wonder which product to choose and why. And yes, I've used DXO's "prime" noise reduction and it's quite good. A big comparison test is really what's called for.
Hmmm, being that I already owned Topaz DeNoise and so qualified for a free upgrade, I downloaded and tried it. I would say that it depends on the image. I tried it on a rather noisy image of a darkened room (see attached image), and it seemed to get rid of some of the paneling detail below the books. Settings are default, without touching the sliders, but even trying the sliders only marginally improved the image. The previous version DeNoise 6, probably made a better fist of the image set to 'Light Jpeg' setting.
Following on from my question concerning external denoising application workflow with Darktable, I put together a play raw to compare how noise reduction compares against ON1 NoNoise, Topaz DeNoise AI and the Profiled Denoise in Darktable.
Thanks for the sample, this is indeed getting interesting! This is now a brand new Adobe DNG 1.7 spec file that has an additional subimage embedded inside (rather than saving it as the primary image after enhancement like the other tools seem to do):
I tend to agree, but it seems that with the advent of AI denoise there seems to be an expectation that photos are supposed to be totally noise free. One of the reasons why I started looking more carefully at NR was because of feedback I received from some judged shows.
Today I updated LR and edited the same pictures with its new denoise function and I have to say LR is better, there might be a little more noise left, but it looks a lot better, very smooth, more natural looking and more details left.
What are the pros and cons to converting to .dng when importing into LightRoom Classic. One pro for me is I can push the changes down to a DNG and then move the raw or open in some other app with the lightroom changes
I have found that I use Photo AI almost exclusively now, but once in a while will use just sharpening or just noise reduction. Good to know that all new development will be contained in Photo AI. I like the ability to adjust the settings!
The level noise and degree of softness in these images is so low I would have never thought to use more than Lightroom for these. It would have been more useful to have chosen images with more noise in my opinion. I use the programs more for problematic images where the sensor is stressed more than would have been in your examples.
If they would fix Photo Ai so you could pick your own setting for the enhancement like lightroom and have use 2 like photo Ai. The photo Ai is a great tool. I like the face part. I still Noise Ai for most noise.
Matt, the latest update to Photo AI (1.3.0) removed the ability to choose which sharpening I wanted: standard, lens blur, or motion blur. As a result, I think the update to this feature was a mistake. There are several feedback comments on this. What is your opinion?
The kingfisher photo looks spectacular when viewed at normal human distance. No photo will ever look good when zoomed that far in but that photo is about perfect and exactly an example of how the software helps.
I have both DeNoise and AI. I also have Sharpen but never been overly happy with that. DeNoise has always been my go to and most of the time it still is. Maybe I need to do a lot more practice with AI. The images I am getting back show horrible sharpening artefacts plus too much blurring of some details. I use micro four third gear which means more noise and as a result loss of detail. And I admit I am sometimes working with a less than perfect image. But in the promo for the software they show fantastic detailed recovery from really bad images and I am not seeing those marvellous results from my processing in AI. I guess I will have to keep trying different things as Topaz has become an essential part of my workflow.
I did ask Topaz about the future of the Photo upgrade plan, due to having a number of Topaz products. No answers yet. I have seen update upon update for Topaz Photo AI. It seems that this very well may be the future?
As a software developer, I would be VERY surprised if the 2 apps used different de-noise engines. I would posit that the differences in your pics are the caused by the differences in the parameters set in each app.
I find Matt that I am finding Photo AI is getting better abd better with each upgrade. However, I still find that by using the three separate programs, Denoise, Sharpen and Gigapixel I am able for really critical work, to refine things to a greater degree than with Photo AI.
I have conducted some comparisons, and I have found that, on some images, Photo AI denoise tends to obliterate very fine detail and textures, even at the lowest setting. On the same image, Topaz DeNoise AI manages to retain the fine detail.
FYI. I have digital scans of old positives that have high levels of noise. Some of these have monochromatic backgrounds. In these instances, I have found that PhotoAI leaves artefacts on the background. In these instances, I have gotten better results using the denoise very noisy options than anything I was able to obtain in Photo AI. In general, I also have obtained good results in Photo AI.
c80f0f1006