PRIME and DeepPRIME continued

So how do the earlier PRIME and new DeepPRIME compare? I've run hundreds of images through both algorithms and compared them side-by-side, and I've found DeepPRIME to offer a pretty noticeable improvement for most images, at least for larger print sizes and when viewed 1:1.

That difference is most noticeable around high-contrast edges and in natural textures like rock, fur, hair and foliage. Compared to PRIME, DeepPRIME tends to make these areas look less splotchy and mottled. Just occasionally though, PRIME still does a better job with things like very fine, repeating low-contrast details like feather patterns that DeepPRIME can lose, so I'm glad to see both options remain available.

Next up, an ISO 12,800 shot from the 20.2 Megapixel Sony RX10 II, again as processed with DeepPRIME.
Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

In terms of noise levels, I've found DeepPRIME to give the appearance of around a three-stop improvement, but it's important to note that it's not a replacement for shooting at a lower sensitivity in the first place, wherever possible.

Its artificial intelligence algorithms can manage to recreate detail quite convincingly, but at the end of the day, it's a clever trick, in much the same way that unsharp masking doesn't actually make your images sharper, but rather fool your eyes into seeing more sharpness with tweaks to contrast.

The crops from the above ISO 25,600 shot by the 32.5 Megapixel Canon 90D show an obvious advantage for DeepPRIME in my son's pupil and iris. Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)
Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

When comparing fine detail against a shot at lower ISO, the improvement is more like a stop or less, and this is especially noticeable in man-made details like fine text where the algorithms can't recreate detail that was irretrievably lost to noise in the original image.

But for many shots, the illusion of that fine detail is all that's needed, and so DeepPRIME can give your shots the appearance of a multiple-stop improvement over HQ denoising, which to my mind already handily beats Adobe's noise reduction.

As mentioned earlier, man-made subjects typically represent the strongest challenge for DeepPRIME. In this ISO 51,200 Canon 90D shot, it nicely resurrects the top of the chain which is all but lost by PRIME, but the fine latticework above looks obviously off.
Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

Performance: DeepPRIME comes at a cost, but no more so than PRIME

The really good news is that DeepPRIME comes at only a small extra cost in speed over PRIME despite its noticeably better results in most cases. To test performance, I assembled a library of 100 Raw images from 18 different camera models, covering everything from entry-level to professional cameras made from 2003 to 2020, and spanning a gamut from 6.3 to 51 Megapixels.

DeepPRIME's advantage tends to be most obvious with natural subjects like fur, hair, foliage and the like. In this ISO 1600 Canon Rebel XTi shot, it does a good job with the fur and the cat's eyes look significantly cleaner, as well.
Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

In all, it made up a whopping 2.9 gigabytes of data, which I ran through all three of PhotoLab's denoising engines multiple times at their defaults. I also passed them through Adobe Camera Raw with lens and other auto corrections enabled, and aiming for a similar output JPEG quality, to allow for a fair comparison.

For reference, I'm comparing DxO PhotoLab Elite 4.1.1 versus Adobe Camera Raw 13.1.1 (which shares algorithms with Lightroom Classic 10.1.1), and my computer is a 2018 Dell XPS 15 9570 laptop running Windows 10 version 1909.

For many subjects, DeepPRIME's clever algorithms work great, but remember, it can't create detail where there isn't any to begin with. This is especially true for fine text; I can't find anywhere in this ISO 12,800 Sony RX10 II shot where I can read text that I couldn't even in the much noisier Adobe version.
Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

To give it its due, Adobe Camera Raw's algorithms (which I've found to be extremely well-optimized for performance) was able to process the full batch in an impressive four minutes, 58 seconds, or just about three seconds per image. With HQ denoising, PhotoLab turned in a time of nine minutes, 38 seconds, or a little under six seconds per image.

Enabling PRIME denoising means a major increase in the processing time to a full 37 minutes, 42 seconds for 100 frames, or a little under 23 seconds per image. But enabling DeepPRIME added only six more seconds to that time, for a total of 37 minutes, 48 seconds. In other words, there's essentially no added penalty for using DeepPRIME over the earlier PRIME engine, and a noticeable improvement in image quality for most images.

It's also worth keeping in mind that PRIME and DeepPRIME may not be appropriate depending on your files and your choice of ISO; in this ISO 400 Canon Rebel T3 shot, Adobe arguably turns in the best performance in terms of holding on to subtle patterns despite the higher levels of noise overall. It's likely that PRIME and DeepPRIME would do better relative to Adobe were this image noisier, taken at a higher ISO value.

Clockwise from top left are 100% crops as processed by Adobe, PhotoLab HQ, PRIME and DeepPRIME. (Click or tap the links for full-res versions.)

And honestly, that added processing time really isn't a big deal unless you're on a tight deadline or planning to process thousands of frames, because it comes after you've finished work and are satisfied that you're ready to render your final images. At that point, you can simply set PhotoLab to work while you grab a coffee and a bite to eat.


Remaining features

The remaining new features in PhotoLab 4 might not be attention-grabbers, but they're very worthwhile too, bringing worthwhile workflow enhancements that users migrating from other pieces of editing software may find familiar.

Selective copy and paste is a great addition

In PhotoLab 4, you can selective copy and paste settings between images, enabling or disabling several dozen different individual adjustments as you choose. These can also be done by groups like Light, Color or Geometry. And as well as global adjustments, you can also choose to copy local adjustments too, which can be especially handy for things like cloning out an annoying speck of dust which remains in the same place across a batch of images.

Batch renaming is handy, but there's room for improvement

PhotoLab 4 also debuts a fairly powerful batch renaming function. This allows you to find and replace a specific string in the filenames, add text before or after the existing filename, or completely replace the existing filename with a custom name, either appending or prepending a counter to it. This counter can have anywhere from one to five digits, and you can choose what you'd like the counter to start from.

I'd like to see this extended further to allow date and time stamps to be added into the mix, as well as other fields drawn from the EXIF headers of the images on which you're working from, but it's a great start!

Watermark your images with ease

Instant watermarking is also new, and can be found in its own panel which, by default, is docked at the right of the screen. You can choose either an image with which to create your watermark or use text with your choice of font, and have control over placement, scale, font formatting, blending method and opacity.

Your watermarks can be saved as presets for quick recall, and can be selectively copied to other images. They can also easily be overridden when exporting images, but not when printing. (I'd like to see DxO add that option to the print dialog.)

Conclusion

Although DxO's new DeepPRIME denoising engine stands out as by far the biggest change in the new release, there are many other welcome changes in DxO PhotoLab 4. I think it's a worthwhile upgrade for existing users, especially those who're still running PhotoLab or PhotoLab 2. And if you're an Optics Pro user, it's really a no-brainer, as PhotoLab 4 is a far more capable offering.

As for those of you using products from Adobe and its rivals, there's still a lot to recommend DxO PhotoLab 4. Firstly, it's nice to be able to forego a subscription, and only pay for upgrades when you feel that newly added features are sufficient to justify the cost. The new DeepPRIME denoising engine is really in a class of its own, and the automatic adjustments are excellent at getting you in the ballpark with a minimum of effort.

DxO PhotoLab 4 is a great, subscription-free alternative to Lightroom if you don't need multi-shot tools like panoramas or HDR

The refined user interface is very accessible and has a relatively shallow learning curve. With the very understandable exception of PRIME / DeepPRIME denoising, editing performance is also great and output performance is pretty good as well.

I've personally used DxO's products in parallel with Adobe's since the Optics Pro days, and with the exception of multi-shot image editing for panoramas, HDR or focus stacking, I think it now offers almost everything I need to replace Lightroom and Photoshop as my go-to digital darkroom application. With a free 30-day trial on offer from the DxO website, I highly recommend giving it a try yourself; chances are good that you'll be able to say the same.

What we like:

  • Superb DeepPRIME denoising blows Adobe out of the water
  • Solid image quality in other respects, too
  • Loads of easy-to-use, reliable automatic corrections
  • Great UI performance and good output speed
  • Deeply customizable and surprisingly approachable interface
  • No subscriptions in sight, upgrade when you choose

What we don't:

  • Colors tend to be a bit muted and exposure corrections subtle by default
  • No support for Fujifilm X-Trans cameras
  • Tends to lag Adobe in speed and breadth of new camera support
  • No support for multi-shot imaging
  • No way to import and tag images in a single step

Who's it for:

Enthusiast and professional photographers with a focus on image quality and ease of use, especially those who shoot in Raw format and at higher sensitivities.