It seems one of the attractions of the babel-image-archives is to find meaningful images within the library.
It seems that meaningful images would typically have less entropy (in terms of the number of different colors and how the position of pixels with similar color). So to increase the likelihood of finding meaningful images, would it be possible to estimate the entropy of an image from the babelia number?
That’s one experiment I would like to do if the source were available. While I’m no statistician, I figure a shoddy attempt could be done by cross-referencing a few points, such as the number of unique colors on a page or neighboring pixels with similar color (given a threshold). Currently, the Image Archive serves an image every 3 seconds, so it’s super slow to make any headway.