2 ms·
The author eyeballed the sources and manually decided on an abstracted color scheme representation. (Remember, no computers in 1902.) You could do the same thi
by jacobolus 11y ago
The author eyeballed the sources and manually decided on an abstracted color scheme representation. (Remember, no computers in 1902.)
You could do the same thing today in a more automatic way by using by using some kind of clustering algorithm on the colors of pixels. I would recommend first applying some kind of bilateral-filter blur, then clustering the pixels (or some random sample of pixels), representing each cluster by its centroid in a roughly perceptually uniform color space (e.g. CIELAB or CIECAM02), and then apportioning squares to each cluster based on the relative number of pixels included.