In terms of differentiating 'interesting' images, would it also be equally as valid to look at other aspects, for example, the color data of the image, instead of just the edges?
@Log Yes, absolutely. Though we'll see later on that overall intensity (i.e., brightness) is the most important factor in terms of human perception.
Can you explain in greater detail on how the norm of the derivative captures the edges?
The concept of taking the norm of an image as how "interesting" the image is reminds me of how the file size of an image in the computer can sometimes depend on its complexity, so images of the same pixel size but different complexities can take vastly different amounts of memory to store. This seems like a more practical use for an image norm.
How often does changing the method of finding out information allow different edge cases to slip through when compared to other ways of taking norms? Ie if we use edges vs a norm for how bright colors are since "interesting" is subjective, but norms are an objective value.