Introduction
There are only two contenders in the noise wars (IMHO) – DXO Photo Lab and Topaz DeNoise/Photo AI. These two have class leading image noise reduction and I have reviewed them numerous times. Towards the end of 2022, both DXO and Topaz issued new versions of their products. DXO released Photo Lab 6 with enhanced noise reduction called DeepPRIME XD. Topaz launched a new product, Photo AI, which includes their latest DeNoise AI technology.
But which is better?
This review examines how these tools cope with a very noisy image, shot on a Canon EOS M50 at 25,000 ISO. The original photo is unusable because of the excessive noise in it.
Download links
To get a full functional trial of these tools, use these links:
Is Noise Really an Issue?
Yes, it is. Even if you own the ‘best’ full frame DSLR/Mirrorless camera, there is a point where images will be noisy. Circumstances may dictate that you increase the ISO so that you can take a blur free photo. And it may then contain noise.
And what of hobbyists and serious non-professional photographers? They may not have the funds for the latest, greatest equipment; or cannot justify the expense. So, they use lesser equipment or mobile phones, and sometimes end up with brilliant pictures that are spoilt by noise. We still need noise reduction software today…
The Original Image
Seen at 100% magnification, this image is a worst-case scenario for noise reduction software. There is both excessive noise but also considerable detail – the teddy’s fur and the red scarf. The challenge for the software is to correct the noise without losing detail. That is quite an ask.
Original Noisy Image
RAW Noise Removal
The best time to remove image noise is during raw conversion. Why? Because image adjustments can introduce artefacts and blemishes. Removing noise from the raw image data cleans it up prior to making other changes, which means you are working with the best possible version of the image.
Topaz Photo AI and DeNoise AI both work with raw images. In their standalone versions, they then output cleaned up DNG versions of your images, which is ideal for processing in a dedicated RAW converter.
DXO is a full-blown RAW converter, and can produce images that are fully processed and ready for printing and display. For this experiment, I only used DXO’s noise removal options. I developed the image twice. Once using the old DeepPRIME function, and once using the new DeepPRIME XD technology. DXO saved these as DNGs, although it also offered TIFF and JPEG.
To provide a further comparison, I used Affinity Photo V1 to develop the RAW image, limiting its processing to noise removal only.
RAW Noise Removal – The Results
Affinity Photo
The best description is – a valiant effort. It’s not bad and the resulting image is usable.
DXO DeepPRIME
Far better than Affinity Photo’s result.
DXO DeepPRIME XD
DeepPRIME XD is a quantum improvement over DeepPRIME. Look at the wall and the bed frame – the improvement is obvious. The teddy bears’ eyes are slightly clearer. There is more detail in the fur. DeepPRIME XD’s result is fabulous.
Topaz Photo AI
In the past, Topaz Labs have lagged slightly behind DXO in the noise removal department. But things change and I would now rate Photo AI’s result as at least as good as DXO’s. Viewed at 200% Topaz Photo AI has kept more detail, especially in the red scarf. Of course, this is splitting hairs. I doubt that at ‘normal’ viewing sizes (on screen or in a print) the differences would be discernible. But I would rate Photo AI’s result as just edging DXO’s – something I’ve seen nothing do before. Wow!
TIFF/JPEG Noise removal
Removing noise from a TIFF or JPEG is much more challenging and DXO has a bigger issue here – its PRIME technology only works with RAW photos and it only offers its High Quality noise reduction for non-RAW images.
Topaz Photo AI and DeNoise AI work on both RAW and developed formats, and this is a tremendous advantage. As the results show, DXO simply cannot compete with Topaz when processing very noisy TIFF and JPEG images, making it the go to tool for mobile phone and non-RAW photographers.
I also tried an old favourite, Photo Ninja, on the noisy TIFF image. I couldn’t try Ninja on the RAW image as it doesn’t understand Canon M50 RAW images and I couldn’t get a decent result from a DNG of the image produced by Adobe’s DNG converter. This is such a shame – Photo Ninja is a great program, but its understanding of modern RAW formats is inadequate…
Photo Ninja’s noise removal algorithm is about the best non-AI function there is. Its results for the TIFF image are far better than DXO’s but are nowhere near as good as Topaz Photo AI.
What Noise Removal software should I get?
If you shoot JPEGs then the decision is simple – get Topaz Photo AI or Topaz DeNoise AI. Topaz Photo AI has some advantages over Noise AI – it contains both Topaz Gigapixel AI (the best image enlargement program around) and Topaz Sharpen AI (the best image sharpener), albeit with less adjustment options than the individual programs. It also features face detection, allowing you to only sharpen the faces in the image and leaving the backgrounds smooth and blurry. As Photo AI is cheaper than the combined price of the three tools it contains, it may be the best option.
If you shoot RAW images, then the choice is harder. Topaz Photo AI edges DXO Photo Lab for noise removal, but by the slimmest of margins. DXO Photo Lab is a fully featured RAW Image Processor and can do far more than remove noise and sharpen images. But I don’t think DXO has enough to produce complete results. My workflow used to be:
- Process in DXO and produce a 16 bit TIFF
- Process in Topaz products to sharpen and enlarge
- Use my Digital Photo Guardian program (currently not available for sale) to produce the final JPEGs for printing or posting to the web/social media.
But now, with very noisy images, my workflow will be:
- Process the RAW image in DXO to use its lens corrections, saving the image as a DNG
- Process the DNG image in Topaz Photo AI and save as a DNG
- Process the DNG in DXO and produce a 16 bit TIFF
- Process in Topaz Photo AI to sharpen and enlarge
- Use my Digital Photo Guardian program (currently not available for sale) to produce the final JPEGs for printing or posting to the web/social media.
Adding two steps to my workflow is not something I do lightly. But I like to get the best possible results and this is the way to achieve this.
My recommended image software toolkit is now:
Digital Asset Management
Exposure X7 or DXO Photo Lab. I actually use Digital Photo Guardian (my own program) but that’s not commercially available.
RAW Converter
Noise Removal/Image Enlargement/Sharpening & Blur correction
Image Editor
Specialist Image Filters
What about Luminar Neo? I’m not up to date with it. I still use its Landscape filters occasionally. But they have gone down the monthly subscription route, which was what made me leave Adobe. Yes, they have a Lifetime License option, but it’s a huge outlay if you include all Expansion Packs and without them, Luminar doesn’t really have an edge. Given Skylum’s habit of relaunching Luminar as a ‘new’ product every year, I wonder how long a ‘lifetime’ license would last and if it would be valid for all future versions? I may well review it if I can get hold of a trial version.
But for the moment, if you’re looking to step away from your current toolset and software subscriptions, then the combination of DXO and Topaz is an unbeatable combination. Affinity Photo is everything you need from a photo editor.
I appreciate the review. It sure helps a non software guy like me figure it out. I do use software but I don’t understand it or how it works. Adobe has a new noise reduction in their Camera Raw and I’m not sure how that would compare. I’m guessing not as well since Adobe lags behind.
Ah yes, Noise Ninja… somewhere I still have a license.
Thanks for this comparison. I am currently a happy user of Affinity Photo, which really is remarkable for the price. Great for retouching. But for noise reduction I am looking for better. The price of additional software, especially considering the hefty upgrade fees each year, is really something.