Previously, I had had used Topaz Labs DeNoise AI with a high ISO Milky Way image, which I wrote about here. Its ability to reduce noise while keeping the stars sharp and intact was impressive. I decided to really challenge it even more at one of my favorite locations nearby, Vasquez Rocks in Agua Dulce.
Vasquez Rocks is a state park north of Los Angeles. It’s a special place, particularly for “Star Trek” lovers. After all, it’s the location of the epic slow-motion battle between Captain Kirk and the Gorn lizard-man.
It’s also a nearby place to photograph the night sky. I decided I would do some comparisons. From the same camera and tripod setup, I wanted to compare two scenarios — a single exposure vs. Topaz DeNoise AI, and a stacked photo vs. Topaz DeNoise AI.
What is stacked?
Many people love to stack photos to reduce the camera’s digital noise. Stacking is when the photographer uses the same camera and tripod setup and the same settings and takes several photos in succession.
The stacking process is based on the median of all the stacked exposures. The final image shows the parts of the exposures that are consistent through each layer, like the stars. However, it removes random elements such as digital noise as well as elements that change from one exposure to the next, such as airplane trails. These random elements are rendered invisible in the final image.
Many people use Photoshop. I use an app called Starry Landscape Stacker (Mac only) to combine the images.
We’ll compare all the photos at 200% zoom. I realize that no one looks at their photos this way, but for the sake of comparison, it makes it easier to see.
Single exposure vs. Topaz DeNoise AI
I wanted to see how well DeNoise AI would work again. I took a single exposure high ISO photo and ran it through DeNoise AI.
Above, I chose the Low Light from the 4-panel AI Model options. Looking at the comparison, you can once again see that DeNoise AI has significantly reduced the noise while keeping the details of the rock and stars intact. I did have several hot pixels, and it did not remove those, but everything else looked great.
Stacked image vs. Topaz DeNoise AI
I often stack my Milky Way photos. I’ve done this for several years to reduce the noise. I used a stack of 15 photos for the comparison. How would DeNoise AI, uh, stack up against this tried-and-true process that night photographers and astronomers use?
Looking carefully at both images, you can see that the stacked image on the right looks cleaner than the other. The stars have shifted because it took the photo in the middle, but regardless, looking at the noise in the rocks and sky, you can see that there is a bit of a difference. Again, do bear in mind that we’re viewing this at 200%, and no one would ever view your photos this microscopically.
If stacking is so clean, why don’t you do that all the time?
It’s a fair question.
Unfortunately, stacking is more involved. If there is a foreground object such as the mountain in this photo, you will possibly need to mask the foreground since the stacking process follows the stars and could render the foreground blurry.
Night photographers encounter this issue to an even greater degree when using a star tracker or Pentax Astrotracer for exposures of a minute or more. Since the camera is tracking the movement of the stars, the foreground will blur. You would then need to take a photo of the foreground separately and blend it back in. In other words, it’s considerably more work!
When is a good time to use DeNoise AI?
Any time! For me, I might want to apply DeNoise AI on a single exposure high ISO photo because it looks quite good and it’s easy and quick to do. DeNoise AI is also excellent for any image with shadows, underexposed photos or low ISO photos.
And for viewing on Instagram or even many prints, the quality of noise reduction here is excellent. Bearing in mind again that we are looking at these photos zoomed in at 200%, DeNoise AI was again impressive.
Above: the image on the left is a single exposure processed with Topaz Labs DeNoise AI. The image on the right is 14 images stacked using Starry Landscape Stacker. Stacking is a commonly used technique among astronomers and night photographers. Here, however, despite this somewhat unfair comparison, DeNoise AI still performs admirably.
Thank you for the articles you have written regarding Topaz DeNoise AI. I have done similar comparisons to Stacking and other noise-reduction software. I have found DeNoise in some cases to be superior to Stacking and in all cases superior to other noise-reduction software. So I have gone to entirely to NeNoise (and all updates) for any low-light environments.
I also use Topaz’s Sharpen AI and GigapixalAI (and all updates.)
I suspect that at some point in time there will be even more advanced such software, Topaz, etal.
Thanks! That’s cool that you’ve found it to be superior to stacking in some cases. Machine learning holds much promise for simplifying our work flow. I’ll try and do some more Denoise AI articles if I can think of any. And yes, I think between Denoise AI, On1, DxO, Photoshop and others, there will be quite a bit of competition and advancements. This is great…as consumers, we all win when this occurs!
Looks like a good tool to have access to.
It’s excellent, and only can get better from here. While it’s hardly foolproof, it works surprisingly well most of the time, and can make high-ISO photos far more pleasing. I haven’t had to use luminosity masks or Layer Masks in conjunction with this so far, but that’s always an option if it mostly gets it right but botches one part.
I have Topaz DeNoise AI. It works great. I will possibly update my copy of On1, supposedly it is as good as Denoise but faster.. we will see. I have read that DXO is the best because it removes noise during the de-mosaic level at the beginning of the raw workflow. Where most denoise is applied at the end. I don’t know if this is true or not I have already spent enough on software, I am not getting DXO. There is agreement by some that say noise should be eliminated at beginning of the workflow prosess. Not at the… Read more »
I think we’re going to see this battle for denoising. And the winner of the battle is….us! I’ve heard it several different ways about which is better for what. My guess is that since there are numerous variables, including which iteration of denoising software, its application, the perception of the user, the type of noise, and more, it may depend. For now, I have Topaz Denoise AI, and I’m generally quite happy with the results. In this article, I compared Topaz Labs Denoise AI to stacking photos. Stacking is utilized by astronomers to reduce noise, so it works pretty well… Read more »
I do a lot of astrophotography with a tracker. Some are 6+ minute exposures and some at 2400 ISO. I have done a lot of testing with different techniques and software. Topaz Denoise AI is so good that I have not done any stacking for some time. I often will shoot a dozen long exposures of a target like a galaxy. I check through the images and discard any where the stars are not round or there are aircraft light trails. I only process the best of all those shots. I use Topaz Denoise AI and Topaz Sharpen AI as… Read more »
That’s impressive. I haven’t done a lot of Milky Way photos in a while, but have been impressed with how Topaz Denoise AI addresses some of my older MW photos. Thanks!
Thanks for this … I certainly was wondering. I’ve used Sequator and Topaz Sharpen both to good effect, but sometimes get confused on which to pick, workflow, etc.
I’m increasingly using topaz denoise on single exposures. I think it performs very well and is genuinely on a par with stacking. My reasoning is that with stacking, factors come into play with many scenes. Branches on trees moving in the breeze and being in f different positions is common. Also low clouds moving through parts of the scene. I also find that starry landscape stacker masks leave a rim light edging to mountains and hills. Besides this, it means I can work more quickly in the field and on a clear night cane come away with multiple compositions and… Read more »
Great points. For very similar reasons, I’ve been doing more single exposures with the stars as pinpoints of light as well.