Adobe Super Resolution Review: A Good Step, Not a Revolution
You’ve probably heard of Adobe’s new Super Resolution feature, which doubles a photo’s linear resolution (quadrupling the total pixels) with better results than any other upsampling algorithm – at least, that’s the claim. I recently got around to testing it, and here’s how it measures up.
Table of Contents
What Is Adobe Super Resolution?
In the 13.2 version of Camera Raw, Adobe added a new option called Super Resolution that can upsample any image to 4× the original number of megapixels – in other words, doubling the width and height of the original image. A 12-megapixel original image would become 48 megapixels; a 48-megapixel original image would become 192.
Rather than using a traditional upsampling algorithm such as Preserve Details or Bicubic Smoother, Super Resolution uses an artificial intelligence upsampling algorithm that Adobe says was trained on “millions of photos” in order to give better results than usual.
How to Use Super Resolution
It’s easy to apply Adobe’s Super Resolution to any raw file that you have. Open the image in Photoshop’s Camera Raw dialog, right-click on the image, and click “enhance”:
The Super Resolution options appear as follows:
It’s a simple dialog, and once the Super Resolution box is checked, Camera Raw will give you an estimate of how long the upsampling will take. Click Enhance, and after some time, the higher resolution image will pop up as a second version of the image:
I’ve highlighted the second image in red in the screenshot above. That’s the high-resolution version; the other one is the original. Make sure to click the second image, or you’ll just keep editing the prior, low-resolution version.
Once you’ve done that, you can edit the photo to your heart’s content in Camera Raw or Photoshop itself.
Applying Super Resolution to JPEGs and TIFFs
Even though Super Resolution is only a feature of Adobe Camera Raw, it’s still possible to apply it to JPEG and TIFF files. Here’s how.
First, in Photoshop’s top menu, go to File > Open.
Then, click on the JPEG or TIFF image you want to edit.
Before clicking “Open,” take a look at the screenshot above, where the bottom menu says “Format.” It’s most likely going to say JPEG by default. After clicking on the image you want to edit, you need to change the Format menu to “Camera Raw”:
Then, when you click to open the image, it will appear in Camera Raw, and you’ll be able to apply Super Resolution just like before. Keep in mind that this doesn’t change the image into a raw filetype like DNG, but simply opens the JPEG or TIFF in Camera Raw.
Note that if you are working with a compressed JPEG file, it’s likely that JPEG compression artifacts will be exaggerated upon applying Super Resolution.
How Good Is Super Resolution?
Everything above is nice to know, but it won’t matter if Super Resolution isn’t any good. So, how does Super Resolution stack up?
The answer is in the title of this review: good, but don’t expect a miracle.
A 12-megapixel image that has been enhanced to 48 megapixels with Super Resolution won’t match an original photo from a 48-megapixel camera, or even get especially close. If that’s what you were hoping for, you’ll need to temper your expectations.
Super Resolution also isn’t drastically better than the Preserve Details 2.0 upsampling algorithm that Photoshop has already had since 2017. Don’t get me wrong – I’ve been impressed by Preserve Details 2.0 for years, and any improvement is welcome. But don’t read too much into headlines claiming that Super Resolution is a revolution. (Say that five times fast!) Instead, it’s a solid advancement to impressive technology that already existed.
Ah, you won’t believe me without tests. Here they are. First, this is the uncropped image I’ll be using. The original is a 47-megapixel image from the Panasonic S1R, and all the crops you’ll see in a moment are from that small red rectangle:
The upsampled images we’ll be comparing are 188-megapixel behemoths – four times the original photo’s resolution (AKA twice the linear resolution). First, acting as a control, is the basic upsampling algorithm of Bicubic Smoother. Click to see full size:
Now let’s take a look at Preserve Details 2.0, the artificial intelligence upsampling algorithm that Photoshop has had since 2017:
And then Super Resolution, the newest of the three methods, and the subject of today’s review:
Super Resolution is indeed better than Preserve Details 2.0. If you’re not seeing it, pay attention to the building on the left, which is crisper in the Super Resolution image. Both of them are clearly better than Bicubic Smoother throughout the image, particularly in areas like the trees and red advertisement on the right, and the same building on the left.
For comparison, here’s how an actual 188-megapixel image looks:
I took this photo using the Panasonic S1R’s sensor-shift mode, so it’s as close as you’ll get to a “real” 188 megapixel shot of this scene. There are so many finer details in this image, and the non-detailed areas have substantially less noise as well. In short, it’s much better than any of the upsampling algorithms.
That’s not exactly a surprise. Adobe would have to do some absurd wizardry behind the scenes in order to double a photo’s linear resolution without losing much apparent image quality. Perhaps they’ll manage something close, one day – after all, the current iteration is better than almost anyone would have expected back in 2010 or so. (Though something similar can be said of Preserve Details 2.0.)
Regardless, I’m happy Adobe has figured out a way to improve their existing upsampling algorithms even further, and hope they can continue down that path. It’s pretty amazing what artificial intelligence algorithms can do today, not just for upsampling, but also for things like noise reduction and fixing motion blur. Any issues I have on this topic are not with Adobe, but with the coverage on various sites that makes Super Resolution seem like never-before-seen technology, when it’s more like a nice iteration of something that’s been around for a few years already.
Adobe Super Resolution vs Topaz Gigapixel AI
Another company that’s using artificial intelligence upsampling algorithms is Topaz, with their Topaz Gigapixel AI software. I figured it would be helpful to add a Super Resolution vs Gigapixel AI comparison so you can see what the state-of-the-art is from more than just Adobe.
Here’s Topaz’s attempt at the same image:
For comparison, here’s a slider between the Adobe Super Resolution image (on the left, “before”) and the Topaz Gigapixel image (on the right, “after”):
(There’s a framing difference because Adobe applied a non-removable lens profile, while Topaz didn’t.)
To me, each image has its pros and cons. The Adobe Super Resolution version on the left doesn’t have as many strange color artifacts or waxy-looking areas. The Topaz image on the right has smaller, finer details overall, as well as less noise and crunchiness. Between the two, I lean toward the Topaz image, but they’re pretty close. As with the other samples in this review, I leave it to you to judge which one you like better and by how much. Personally, given that Topaz Gigapixel AI costs $80, and I already have the Adobe bundle, I’m sticking with Super Resolution.
With Super Resolution, Adobe has done a good job building upon their existing Preserve Details 2.0 algorithm to allow better upsampling than before. Still, it’s arguably a bit behind the results of Topaz Gigapixel AI, and all of these upsampling algorithms are, unsurprisingly, far behind an original, high-resolution image. So, I take issue with reviews that claim Super Resolution drastically reshapes… well, much of anything.
Not that it needs to. Any improvement to a good foundation is welcome, and Super Resolution is indeed better than Adobe’s prior best upsampling algorithm. I’m hopeful that they can push the envelope even further in the future. Upsampling isn’t something that most photographers need to do very often, but when it’s necessary, it can make or break a print. Adobe is pushing things in the right direction with Super Resolution, and even though I’d say the hype grew a bit beyond the reality, the reality is still very good.