I sit a few buildings down from Cindy's office and have also been studying colormaps quite a bit in the last 3 years. Its interesting how many plotting packages get this wrong but are finally catching up.
I switched from Matlab to Python years ago and was sad to see pyplot using the default rainbow palette still. However, there was some good work done by Chris Beaumont to improve the plot quality. See: http://plotornot.chrisbeaumont.org/ You can easily import these styles into matplotlib using rcparams.
Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab.
I want to talk about the Luv Lab colorspace. There are several places on the net (even in the literature) that are wrong about these colorspaces saying Lab is for emissive displays and Luv is for reflected light. This is actually not true. (If anything it is reversed). See: https://groups.google.com/d/msg/scikit-image/DIRaSXJoEes/2jD... and Berns reference.
The interesting with colorspaces (and colormaps thereof) is that working in a perceptual space like Luv/Lab is yields a non-linear (and non-convex) gamut in the sRGB space used by most monitors. There is more "headroom" in the magenta hue of colors than say green. However, you have to then look at monitor output as a function of hue and human sensitivity -- with a red object and blue object with the same reflectance under the same illumination, the red object will appear darker to humans. So there are many transfer functions at work here which makes the problem challenging in picking the right colormap that is perceptually uniform, has the maximum number of perceived differences, and has the appropriate number of hues for best represeting your dataset.
> Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab.
And Matplotlib is switching to a perceptually-linear default colormap called "viridis" with the 2.0 release (due early this month): http://matplotlib.org/style_changes.html
I’m glad that people are finally catching up to Cynthia Brewer (who has been discussing the issue for 20+ years) and using reasonable color models to think about data displays.
It would be nice if color pickers, color manipulation tools in graphics software, etc. would also switch to more perceptually relevant models.
If anyone wants to play with the CIECAM02 model discussed in the talk you linked, I made an implementation in Javascript several years back, https://github.com/jrus/chromatist/
Here's the ColorBrewer tool: http://colorbrewer2.org/
I never thought about what a PhD in cartography would look like. Nice to learn more about the thought being put into maps.