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.
With some detailed testing, I can say that Topaz DeNoise AI works better than any similar software I’ve ever tried.
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.
Here’s an extreme enlargement of part of a photo taken just after sunset. You can see the before and after noise reduction and the preservation of details.
Every photo is different, and your adjustments will vary from image to image. It’s not automatic. But it’s also quick and easy to make changes and see the results.
I tried DeNoise AI as a standalone app and inside Photoshop as a plugin. Results were identical, as expected.
When installing, I didn’t see any offer to add DeNoise AI as an Adobe plugin as many installers do. Yet, when I opened Photoshop, it was there as a plugin.
I did write to Topaz support to ask about if it was usable as a plugin, and I was surprised to get an email back within 10 minutes. Impressive support.
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,
What I Liked
- Very effective at noise reduction without losing detail
- Easy to use: you can adjust in (almost) real-time to get the results you want
- Reasonably priced, especially with introductory discount
- Use as a standalone app or Photoshop/Lightroom plugin
What Could Be Better
- Processing can be slow: up to a minute or more on large raw files or 16 bit TIFF images
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.