> Does super-resolution add real information to the image? Can it be used for real world analytical applications? [...] The answers to these questions are resou

> Does super-resolution add real information to the image? Can it be used for real world analytical applications? [...] The answers to these questions are resounding YES I have no idea whether it might be useful for real-world analytical applications. But I am quite sure that it is impossible - metaphysically impossible in an information theoretic sense - for a filter to "add information to the image", in any sensible interpretation of that phrase. Where does that information come from? > into the corresponding most likely image of the desired high resolution according to the acquired prior knowledge That's like me taking the first image in the example sequence [1], recognizing it's an airport, and then drawing from memory, what an airport might look like. If I know those grey blobs are airplanes and the black strips are tarmac, I might come out with something not unlike the AI produced. But that's what the result is - a statistically plausible interpretation. (In the AI's case, an interpretation masquerading as if it were an actual photo.) Have I "added information" to the original photograph by interpreting it in a certain way? I would say no. [1] https://miro.medium.com/v2/resize:fit:720/format:webp/0*gUcvD5vkB3r7y6jO.png https://miro.medium.com/v2/resize:fit:720/format:webp/0*gUcv...