Artificial Intelligence and Machine Learning are everywhere in photography now. I recently wrote about whether we should be worried about AI in photography. My conclusion is that we shouldn’t. Today I’m demonstrating the power of AI in the form of a battle, putting the leading noise removal software head-to-head. I’ve searched through my raw images to find a particularly noisy shot to work with. Here’s what we’re working with:
I’ll quickly explain what’s going on here. This is a shot from the shores of Jökulsárlón in Iceland. It’s better known as the Glacier Lagoon. I took this shot last winter on a moonless night. When I’m looking for the northern lights, I take quick shots to help me compose, which is what this is. I’ve selected this image because it was taken in the dark, and that’s where we often encounter noise in our photos.
The noisy photo information
- Shot on a Nikon D810
- ISO 8000
- Aperture: f/2.8
- Shutter speed: 0.6 seconds
You can see just how much noise there is right here when I zoom in:
The Contenders for noise removal
In this battle, I’ll be using the following software and features:
I’m opening the raw photo directly into each piece of software without any prior edits. Where there’s an automatic function available, that’s what I’ll be using.
I’ll go through each piece of software now and have a complete side-by-side at the end at 100%.
Adobe Sensei drives some awesome Neural Filters, but their Noise Reduction Neural Filter isn’t available yet. We’ll have it covered as soon as this drops, but I’m using the Reduce Noise Filter for now. When we see the full image, the difference is barely noticeable. I’m a big fan of Adobe, but they’re falling behind when it comes to noise reduction.
The noise reduction features currently fall within the Filters section of Adobe Photoshop. They are all manual, giving us options to change by means of slider adjustments. All the other software I used in this comparison was far simpler and more effective. We’ll see what happens when the Neural Filter lands.
Skylum Luminar Neo
The Noiseless RAW feature within Luminar Neo recommended that I use the ‘High’ preset. The noise reduction took about 45 seconds, and the result was pretty good. The expectation in most noise removal is that the details will be softened and the pixels will appear to have been ‘spread out.’ This is not the case at all with Luminar Neo.
The before and after images here show some difference at the 100% scale. The difference is far more obvious when we’re zoomed in tight (at the bottom of the page). The Noiseless AI feature of Luminar is part of their Extensions Pack. Their subscription model makes the Extensions available to users as soon as they’re ready to go in their $119 annual plan.
Topaz DeNoise AI
Topaz Labs are probably the most talked about in this line-up. Their ads appear all over social feeds, and with good reason. Their AI software is powerful and productive. This DeNoise process was easy and intuitive. The software took care of everything for me and churned out a great result, taking a minute and one second to complete the task. This software does what it says on the tin, with no frills.
I prefer this workspace. The side-by-side view reveals a before-and-after, and the sliders and buttons are easy to see and read. The preview took a while to hit 100%.
ON1 NoNoise AI
ON1 NoNoise AI was almost instant, taking three seconds to export the JPEG after hitting the ‘Done’ button. It had no problems in processing the RAW file, and I didn’t need to make any adjustments to the sliders. The preview was instant, whereas the Topaz preview took about twenty seconds to appear. This means a faster and more efficient workflow. Great job, ON1.
Another feature present in the ON1 NoNoise AI is a Tack Sharp AI option alongside NoNoise AI. This sharpening tidies up a lot of the fuzziness that usually results in images processed by noise reduction software.
Results (click for high res)
If I had to pick a winner, I would be torn between Luminar Noiseless AI and ON1 NoNoise AI 2023. Topaz isn’t bad, though, so it’s really tough. The truth of the matter here is that each piece of software is evolving. That’s the nature of Machine Learning – the Artificial Intelligence develops over time.
What is currently the best option may not be in a few months’ time. The same can be said for the images we put through the software. High ISO images and low light images can produce different results, and the camera and sensor also make a difference. Each piece of software is essentially an ecosystem. Just like when you buy into a camera brand, you are basically “stuck” with it. The same applies here. The good news is that Topaz, Skylum, and ON1 all offer a free trial, so you can test each piece of software on our own images to see which works best.
Final Answer for this battle:
- ON1 NoNoise AI scrapes the lead
- Luminar Noiseless AI is a very close second
- Topaz Denoise AI is a very close third
- That said, I’d use any of them at any time – they’re all really good. Each is constantly evolving, and their respective product ranges are constantly growing as AI comes to the forefront in photography post-processing.