Implement paint mixing simulation

[thomasp85]

https://stackoverflow.com/questions/1351442/is-there-an-algorithm-for-color-mixing-that-works-like-mixing-real-colors

[dylanbeaudette]

Some ideas related to reflectance curves for Munsell chips or sRGB values

• http://www.munsellcolourscienceforpainters.com/MunsellResources/SpectralReflectancesOf2007MunsellBookOfColorGlossy.txt
• http://scottburns.us/rec-2020-rgb-to-spectrum-conversion-for-reflectances/
• http://scottburns.us/fast-rgb-to-spectrum-conversion-for-reflectances/
• http://scottburns.us/subtractive-color-mixture/

[teunbrand]

Aside from the elaborate mixing mentioned in the posts above, it would also be nice to have a straightforward colour mixing/interpolation function (in some space) without the need for simulation. This can already achieved manually, but it might be useful for r-lib/scales#423 to have that extra farver magic.

[dylanbeaudette]

I've been thinking about wrapping this code into an R package, unless @thomasp85 is interested in putting it here...?

https://github.com/scrtwpns/mixbox

The mixbox approach is really interesting because it can closely approximate a subtractive mixture without any simulation.

Currently, there is support for the simulation of subtractive color mixing the the aqp package, but only for colors referenced in the Munsell system and having reflectance spectra. I plan to add more reference spectra in the future. Even so, simulation is expensive and the R implementation isn't very fast.

[thomasp85]

I'm very interested in having it here. One thing to be super careful about is the licensing of the various pigment mixing approaches as I remember them to be very restrictive. It seems mixbox is CC-NC which seems fine but better check. Often there is copyright on the reflectance data

[dylanbeaudette]

Great! I've had some fun conversations with the mixbox authors and they have been very supportive of its (non-commercial) use. I'd be happy to help with testing and or documentation. Compiled code isn't my specialty.

I'm planning to further develop the Munsell / reference spectra based mixing in aqp, as it is mostly concerned with "small mixes". In addition, I'm planning to add diffuse spectra for a simulation closer to that outlined by Kubelka–Munk.