It' the individual frames that are compressed, essentially the video is unpacked and detail is culled from averages across multiple other frames beside it. So if the top of the video, for example the sky, doesn't change then that part will be kept static.
It's not so much properties about the video, but properties about each frame. I can take a 1080p image and blow it up to 8K in GIMP, but it's got the same detail as a 1080p image.