heatmap color scales

[gusl]

I'm programming something that makes color plots.

If I interpolate linearly between blue and red, I end up with different shades of blue, red and purple.
But most heatmaps also have greens, yellows and oranges, which I think do more good than harm.

Is there a standard path through the RGB color cube?

Comments

[http://users.livejournal.com/_tove/]

This is often done with spectral colors.

[gfish.livejournal.com]

I recommend against using RGB, as it really doesn't represent how the human eye sees colors. For a quick and easy fix, try using HSV. If you vary the hue value, you'll get something a lot more linearly interpretable.

If you can quantify the ranges down to around ~10 bucket, then I highly recommend looking at the color schemes listed on ColorBrewer. It's what I used in my big (sadly offline) GIS startup project, and they're simply gorgeous.

[(Anonymous)]

There's a bunch of standard paths... here's some:

http://www.scipy.org/Cookbook/Matplotlib/Show_colormaps

:)
Terry

[(Anonymous)]

Try http://cran.r-project.org/web/packages/colorspace/index.html

That's an R package that'll generate the colors you want. Also links to a paper on the theory.

[gustavolacerda.livejournal.com]

Thanks. Who is this?

[gustavolacerda.livejournal.com]

Thanks.

I would love to see actual curves showing how much each frequency activates each cone-type (in a typical human).
I suspect that they are roughly Gaussian-shaped and overlapping (otherwise the rainbow would have some colorless spots between red and blue).

I suspect the RGB spec is based on our cones, in such a way that the RGB colors are produced by mixtures of the 3 frequencies.

[http://users.livejournal.com/_tove/]

If I understand correctly what you're asking for, Wikipedia's got you covered there, too.