more average colours

i finished the thing...

added a conversion to lab space (with skimage.color.rgb2lab) and did the clustering / averaging in that space
convert back to rgb (skimage.color.lab2rgb)
- this was still a bit too dark
adjust the exposure with skimage.exposure.adjust_log^[1] (using the default parameters).
created a colour map with the help of ListedColourmap and BoundaryNorm. the latter of which is new to me, along with the spacing="proportional" argument that allows the height of each block of colour in the colour bar to be proportional^[2] to the range it spans.

and so. here’s Ρόδος (fig 1)

three maps: satellite, DEM, filled contours with colours from satellite — Fig 1: sentinel-2 image (2023-09-14) of the a little bit of the greek island of Pόδος; digital elevation model (copernicus global dem) coloured with `gist_earth`; tanaka contours with levels coloured according to their average colour in the S2 image.

the code for this is ok. not super fragile. but far from here’s-a-nice-easy-to-use-package-that-any-schmuck-can-use. sorry. working out how to do such a thing is on the (back of my mind) to-do list.

will i ever use this in a ✨science✨ context?

no. of course not.

will i ever use it?

no. probably not.

does this mean this was a largely pointless exercise?

no. here’s another one (fig 2).

filled contours of an icelandic fjord, the colours mimic what would be seen from space — Fig 2: seydisfjordur and mjóifjörður, iceland. colours from sentinel-2. elevation data from copernicus global dem

an update

seasonal palettes

i had the thought when first doing this, but didn’t bother to actually do it. but that wasn’t good enough. so, here (fig 3) is the same area, with the same elevation bins, coloured in with the colours of late winter and summer.

two images showing terrain with filled contours, the top image (titled: March) is more brown, and the snow extends lower down the valley; the bottom image (titled: August) is greener with a higher snow line — Fig 3: dolomiti di brenta, in march (top) and august (bottom). colours from sentinel-2. elevation data from copernicus global dem.

another idea

to identify individual water bodies^[3] and get the average colour for those separately.

footnotes

also tried adjust_gamma and adjust_sigmoid. at the (many) few locations i tried adjust_log helped the most. why? dunno. ↩︎
quelle surprise ↩︎
with (a) an ndwi threshold & (b) an area (minimum number of connected pixels) threshold ↩︎

more average colours

# an update

# seasonal palettes

# another idea

# footnotes

an update

seasonal palettes

another idea

footnotes