Super-resolution is a technology that adds more detail to an image in post-processing than is originally available in digital form. Simply put, they are methods that increase the resolution of an image and do more than just interpolate between known pixels to get the "more pixels". A super-resolution (SR) application must therefore coherently add details.
With the advent of Deep Learning, superresolution has received a real boost, as AI algorithms are very good at recognising and adding to objects. For example, if an AI has seen millions of eyes from different angles and in diverse lighting situations, it can then add learned details to each eye in an image. However, this is just as true for branches, cars or fur structures. The "art" of creating a good super-resolution algorithm consists primarily of huge collections of photos with which to train the algorithm. And since Adobe has access to huge stock footage mountains, in our eyes a Super Resolution solution has been due for a long time.
And now the ACR team has finally integrated an SR model into Photoshop. That is, more precisely, released in Camera RAW 13.2, which, among other things, comes with the new, native Apple M1 version of Photoshop.
For moving images, by the way, you still have to wait a while for reliable Super Resolution applications. Because which details are added and how exactly depends strongly on the original image. For example, if you have a fur structure that changes between frames due to a movement in the structure, the detail addition over time is not necessarily conclusive. Instead, in many cases you still get "jumping patterns" that change drastically abruptly from one frame to the next.
But here too, research is already in full swing and it is only a matter of time before reliable super-resolution is found for problem-free, daily use in many video applications. more infos at bei blog.adobe.com