Digital Light & Color

Does the method described here make sense to anybody?

https://www.dpreview.com/news/050146951 ... ve-results

It appears to be a variation on some of the AI-based techniques already available -- in other words, it depends on training the algorithm against a large set of sample images. When upsizing, it essentially replaces the missing information with details selected from images on which it was trained.

You just know someone's going to train the AI on a bunch of kitten images.

Another, longer article from dpReview on the topic.

https://www.dpreview.com/articles/50760 ... fic_source