JGoldstone
commented
1 year ago I spent some more time thinking about this today and looked at what is published by ISO. The core document is ISO 17972-1:2015, and there are three related documents covering extensions (CustomResource elements) for scanner targets, output device targets and spot color characterization data. I was evidently wrong about CxF being up for revision this year; that's not true of 17972-1. (It is true for 17972-4, which is what I might have seen earlier.)
The four-part breakdown with staggered release years really underscores the graphic arts applications that ISO CxF targets. We are going to be 0.001 ppm in one drop in just one bucket of CxF's target audience. Our best hope is probably that in broadening scope to the industries served by Colour, we can develop the equivalent of a -5 document for our industry. (I have asked Tom Lianza of Portrait Displays if he knows anyone on ISO TC 130, most helpfully anyone from ANSI who's representing the US there.)
Anyway. We should probably buy a copy of the base document and at least one related document (I'd suggest the -4 document covering spot color spectral data). It's kind of painful to pay those kind of prices, but maybe there's another way (do university libraries, e.g. at MIT or RIT, have standards access?). If nothing else we will see how the modular construction of CxF is exploited in a real-world application.
CxF gets reviewed every five years. The last review was in 2018, so it's up for review this year. Should we try and participate by asking for things we think will keep it modern, e.g. support for the CIE 2015 Standard Observers? If we do that, we should probably go into it with a copy of the genuine, official ISO spec in our hands (which costs CHF 92).
Another way we could improve CxF would be to make it more friendly for color samples that were born-in-the-metaverse, i.e. were CG from start to finish. That's not a business that X-Rite knows all that well, or at least, it used to be outside their make-serious-money business (like, e.g. QC on auto paint).
What about spectral upsampling? CxF is 15 years old, and spectral upsampling was uncommon then. Was there an assumption that if you had tristimulus data, you could not take them and get back spectral data? How can one document using today's CxF that one took some RGB colorimetric data for an illuminated object, and maybe some information about the illuminant hitting that object, and used some well-known or some proprietary [perhaps parameterized] algorithm to reconstruct reflectance spectra? Can the currently documented CxF cope with this?
Will it be straightforward to model emissive devices with five or more primaries? ARRI Orbiter lights, for example, have six primaries.
In general I think we should ask ourselves, in the types of work with which we are all engaged, where would today's CxF not be a solution?