Update: These colormaps have been merged into the development version of Matplotlib, all of them will be included in matplotlib 1.5, and "option D" (now called "viridis") will be the new default colormap in matplotlib 2.0. If you just want the colormaps themselves, they're available in a single file here under a CC0 "no rights reserved" license. Third parties have also made viridis available in R and Matlab and JavaScript / D3. Below is the talk presented at SciPy2015 that outlines the whole story.
The rest of this page is preserved for historical reference. If you have comments, concerns, or questions about the new colormaps, please get in touch via the matplotlib mailing list.
This page gives an overview of the colormaps we (= Stéfan van der Walt and Nathaniel Smith) have designed as potential replacements for matplotlib's default, jet.
First for comparison we show what several well-known colormaps look like using a visualization tool we developed for assessing colormap quality, and then give 3 4 new colormaps that we've designed. We'd like feedback on which one you like best!
Please leave comments on this thread or email us directly (update: see above).
To start with, here's jet:
You can click on the figure to expand it. The annotated parts are:
A) From top left to bottom right: the colormap, its grayscale version, perceptual deltas and perceptual lightness deltas. Thereafter follows four different simulations of color blindness.
A "perceptually uniform" colormap is one for which the "perceptual deltas" plot makes a simple horizontal line. (This is essentially the derivative of the colormap in perceptual space with respect to the data. We want our colormap to have the property that if your data goes from 0.1 to 0.2, this should create about the same perceptual change as if your data goes from 0.8 to 0.9. For color geeks: we're using CAM02-UCS as our model of perceptual distance.)
You can see that jet is not perceptually uniform.
The plot on the right is similar, but shows how perceptually uniform the colormap is if you print it in black-and-white. We again want this to be a flat horizontal line.
Of the 4 colorblind simulations below, the upper-left one -- "Moderate deuteranomaly" -- represents by far the most common form. It affects something like 5% of white men. (Well, this is a slight simplification because some people will have somewhat stronger or weaker colorblindness, and this just picks a single point in the middle of the scale. But that's good enough for present purposes.)
B) A 3-dimensional visualization of the colormap spline through CAM02-UCS color space. This is more interesting in the interactive version. The little dots are equally spaced in the data space, so "perceptually uniform" in this plot means that the dots should be equally spaced in the visualization. In the interactive version (see below) you can click on the "Toggle gamut" button to see an outline of which colors are actually possible in sRGB.
C) Example images colored with the test colormap (left column) and how it would be perceived by readers with the most common form of color blindness (right column).
This is MATLAB(R)'s new default colormap.
