Topaz DeNoise AI versus Lightroom
I decided to do some detailed comparisons between Lightroom and Topaz DeNoise AI. The Lightroom product does a pretty good job at sharpening and getting rid of noise, but DeNoise AI is just a little better.
I use DeNoise AI for a couple of reasons. Sure, I use it to tame noisy photos, but I mainly rely on it to sharpen details. The artifical intelligence built into the Topaz products is just superior to the conventional pixel-based image processing logic that is used by Lightroom.
In the sample shots that follow, I always begin with raw-format files. This way, I’m not depending on the camera to “clean up” the image data in any way.
I’m not a fan of using high ISO, because it destroys resolution and loses too much dynamic range. There are times, however, when you have to crank up ISO or else lose the shot.
Topaz Severe Noise algorithm at work on ISO 25,600 shot
The shot above shows the before-and-after results of processing a shot taken at ISO 25,600. I never willingly go this high in ISO unless I’m desperate. The resolution is largely destroyed, and the dynamic range takes a huge hit. If I want to totally rid luminance noise at super-high ISOs, then this is the algorithm that I use.
I used the Nikon D850 to take this shot of oranges. Its sensor is still near state-of-the-art, but even this camera can’t make ISO 25,600 look acceptable to my eye.
Lightroom at work
I processed the same ISO 25,600 shot in Lightroom. The results are actually pretty similar to what Topaz did. The noise is gone, but fine details have been sacrificed to rid that noise. So far, I’d have to say that there isn’t a compelling reason to use the Topaz product instead of the Lightroom product.
Topaz Clear algorithm at work
If I process the same shot using the Topaz Clear algorithm, the fine details look better. A tiny amount of luminance noise is left behind, but the overall shot is sharper. I personally prefer this compromise over total elimination of the luminance grain.
Much of my photography involves long focal lengths around 600mm and motion-stopping shutter speeds in the range of 1/2000 through 1/4000. This generally forces me to use ISOs up to 6400. My cameras are able to retain an acceptable amount of resolution and dynamic range at these ISOs, but only with careful post-processing.
Before I discovered Topaz DeNoise AI, my upper ISO limit was 3200. I feel that its AI algorithms have allowed me to get about one more stop, or ISO 6400, before the shot quality drops too low for my taste.
D500 with 600mm at f/8 1/4000s ISO 2,500
The shot above was shot with a Nikon D500, using an ISO of 2,500. This is much more typical of what kind of ISO is needed with long focal lengths. I’m showing a pixel-level view of a photo processed using Lightroom. The shot is a very distant pine tree that hasn’t been affected too much by atmospheric turbulence.
At this viewing scale, the shot looks quite sharp, and you can still make out the pine needles, even though the tree was several hundred meters away.
Same photo processed with Topaz Clear algorithm
At the same pixel-level view, the Topaz-processed shot shown above looks about the same as the photo processed using Lightroom.
Edge haloes seen in Lightroom processing
If you look more carefully, however, there are edge haloes that can be seen in the Lightroom-processed shot. I drew arrows to show edges that demonstrate these haloes more clearly. The haloes aren’t gross, but they are definitely there.
If you don’t sharpen as much in Lightroom, the haloes will decrease, but then the details start to turn into mush. There’s always a balance between sharpening enough without having the edge haloes become obvious.
No edge haloes in Topaz processing
If you look in the same locations as the shot processed in Topaz, the edge haloes are gone. Also, you see a few more pine needles that don’t seem to even exist in the Lightroom version.
Somehow, the Topaz algorithms sharpen but don’t generate any edge haloes in the process. This seems to be the primary difference between Topaz and other photo editors that rely on conventional pixel-based enhancement logic.
Granted, the Topaz results aren’t a huge difference from Lightroom, but the subtle quality differences are definitely there. I have to mention that it will definitely add more time and effort to use Topaz in your editing workflow, but I think it’s worth it.
Someday, all image editors will be forced to adopt artificial intelligence in their sharpening and noise elimination algorithms. Start saving up to buy the new computers that have the necessary processing power (e.g. GPU) to keep up.