5 ms·
It seems FTA that the ‘average’ color is found by resizing the color information within the bounds in question to 1:1 and then using that color to fill the boun
by doctorhandshake 5y ago
It seems FTA that the ‘average’ color is found by resizing the color information within the bounds in question to 1:1 and then using that color to fill the bounds. I’m not sure but that seems to be an important algorithmic element of this process that is almost completely delegated to a side effect of an image library and not really discussed in the context of this write up.
- lgeorget 5y agoI also routinely make visualizations of geographic and weather data (including satellite imagery) as part of my job and it doesn't surprise me. Once you've 1) acquired the data you need (in this case, we're talking about images of countries without clouds and corrected for exposure across the Earth) and 2) fixed your data (because it's never in the shape you want, or the map projection you need), the core of the algorithm can seem pretty trivial sometimes, especially because R and GDAL come with a very large set of ready-to-use algorithms (and I'll never thank the creators of their respective libs enough for that).
- pieterhg 5y agoFTA?
- taejo 5y agoFrom The Article
- aardvark179 5y agoThere will be a considerable amount of averaging and re sampling before the final resize. Depending on shape of the region and how samples outside it are handled you may see considerable variation in the results. What I would be interested in is how much the colour would differ if another colour space were used. I wouldn’t expect a huge change in the imital re projection as adjacent samples are likely to be close in whatever colour space is used, but I think it might change some of the average colours.
- maxerickson 5y agoThe color is seasonal also. Michigan is green in the summer, loses that in the fall, gets whiter in the winter.
- tomsmeding 5y agoI think that most places on Earth get whiter in winter. ;)