Jon Devo column
Jon Devo tries out Adobe’s new technology for taking your photos from low-res to high-res
How Adobe’s Super Resolution could end the megapixel arms race
Enlarging images through artificial intelligence isn’t new, but recent advancements in computer learning could transform the landscape for ultra-high-resolution photography. Basic implementations of these tools attempt to preserve line detail, while increasing resolution by as much as 400%, filling in the space between pixels where edges aren’t detected. It’s a clever but rudimentary approach that can introduce artifacts.
However, the technology has advanced to the point where computer algorithms can ‘see’ what is in the image. Based on context, lossless AI-upscaling can detect where the lines and detail should be as the pixel count is expanded.
At present, upscaling can only enhance what’s in the original image, so don’t expect to produce detail that was not there in the first place. But if you’ve got a good starting point, the results can be astonishing.
Imagine revisiting your archive and being able to increase image resolution by 400%. My early days in digital photography were spent using cameras like the Canon 40D, the Nikon D90 and the Sony DSC-R1. Having more than the 10-12 megapixels they offered for printing or increased cropping flexibility would have been amazing back then.
This technology also represents a huge boost for people who can’t afford the best high-resolution cameras. While some cameras offer pixel-shift high-resolution modes, which extend their megapixel capabilities, the slight drawback is that they require you to keep the camera still during shooting. This is because the sensor moves to create a large composite image. AI-upscaling only needs one good shot.
Adobe is the latest software giant to develop a dedicated upscaling tool. You open your image – preferably a raw file – in the latest version of Camera Raw. Then you right-click the image and select Enhance > Super Resolution. Depending on your system, it will take
a short time to produce an upscaled file, which will be in the DNG format. The great thing about this is that the new file retains the editing adjustability of the original raw file. [Alsoseep100.]
Looking at the results, I was blown away by the quality of enhancement, especially when working with Raw Mosaic files from Bayer sensor cameras and Fujifilm X-Trans sensors. This is because Super Resolution combines with Adobe’s Raw Details tool to improve colour reproduction and reduce artifacts, while retaining edge integrity during scaling. In many of the examples I tested, the output wasn’t just four times larger: the images had reduced noise.
There are already free tools online that offer comparable results to Adobe’s – but with a market leader incorporating this technology, it’s only going to get better. The megapixel race could soon be done and dusted for good, despite mobile phone makers seemingly desperate to prolong it. Instead, sensor manufacturers could shift their focus back to smallerresolution sensors with larger pixels. If a low-light king like the Sony A7S can be used to produce 48-megapixel stills, who needs 50MP sensors?
“With a good starting point, the results can be astonishing”