CIE 2015 CMFs... what would it really take for our entire industry to adopt them?

Our entire industry’s colorimetry sits upon a ~90 year old house of cards: CIE 1931 CMFs.

What would it ACTUALLY take for all of us to adopt the usage of the CIE 2015 CMFs, and the resultant UCS chromaticity coordinates (s, t), to redefine EVERYTHING, since they’re not linearly translatable from CIE 1931…?

The scientific and practical reasons to “update” have completely convinced me. BUT, the practical implications of adoption (rather, the sheer gargantuan size of the amount of roadblocks) seem impossibly staggering to me to even fathom…

Hi,

Mostly everything is built on-top of the CIE 1931 2 Degree Standard Observer, it is a case where we cannot change easily the foundations without breaking the house. One of the first problems being that pretty much none of the RGB colourspaces is spectrally defined.

Cheers,

Thomas

Exactly. I sat there and thought… what would have to change…?

LITERALLY EVERYTHING.

haha.

1 Like

Cameras manufacturer can run their optimisation with the new observer. So they could update their IDTs…
Also for cameras a generic observer bridge might be good enough to begin with. (You still grade the images)

CMS could be used with little modification.
Main work would be to offer compatibility modes. Or to maintain both pathways in parallel.

Rec 2020 is already spectrally defined.
If SMPTE or ITU would make an attempt to define 709 and P3 spectrally we could switch.

It is that easy actually.

3 Likes

Which optimization space do they use for the objective function though? The perceptually uniform spaces are pretty much all based on the CIE 1931 2 Degree Standard Observer.

I might sound pessimistic but getting them to do that is the hard part!

1 Like

True point, also I believe it is much easier to find better UCS with a 2006 LMS observer than with a 1931 one.

I think what we would need is:

  1. a sensible suggestion
  2. a compelling demo showing the advantages

With these two you attend a few conferences and trade shows.

Sounds like a good research area for academia.

1 Like

Tim,
The biggest problem will be SMPTE. There have been numerous changes to CMF that are all more applicable to motion picture applications at this point, but SMPTE specs are in 1931 CIE. Not meeting those specifications means that you are not meeting specification.

There isn’t any reason why new specifications can’t be defined in 2015 while older ones stay in 1931.

1 Like

I would disagree on that.
A colour space is fundamentally defined against an observer. And you cannot trivially convert between two colour space referenced against two different observers.

3 Likes

Converting between color matching functions is a fundamental function of color science.

Even then that isn’t what I was proposing. New standards could be defined with a new color matching function while older ones are left alone. As the industry is veering into more esoteric imaging technology the 1931 observer functions are being stretched to thin. For nearly a century they were ‘good enough’ that they didn’t need to be changed. There isn’t much point in re-specifying itu-601. Rec2020 could use an update as display industries move to narrow band primaries for direct view LED walls as an example.

1 Like

If the two observers do not share the same metamers you cannot convert between the two in a general sense.

How would you render an image to both old and new standard at the same time?

I think they only way to achieve this is to define the old displays in terms of the new observer, hence the need for a spectrally defined display standards.

2 Likes

Agreed! This is the only practical and clean way to do that. Finding an invertible function that maps between the two observers is not trivial, some areas at both ends of the sensitivity space are very different.

1 Like

If the two observers do not share the same metamers you cannot convert between the two in a general sense.

The basis the formation of CMF’s is the correlation of observers. Every single observer will have their own CMFs.

The fitness function for the evaluation older standards would be 1931CMF based results while the newer standards use the new functions.

Even if the older functions would be redesigned into the new CMF’s that would actually be a new standard as opposed to the update of an old one.

Pretty specifically the hope is that there is a strong difference between the new and old observer results in certain cases. The 1931xy coordinates have significant visual differences while producing identical chromaticity coordinates.

The industry was faced with the failure of of the 1931 functions when the first Sony BVMx300’s were implemented and compared to BVM Crts. The same thing happened again when high gamut digital projection was introduced.

1 Like

Absolutely 1931 is a problem no one is arguing about that.

Sony could minimise the issue for a particular display because they hat the spectral data of that particular display.

The point Thomas and I am making is that you could switch the observer trivially if all the display standards would be spectrally defined.

1 Like

Sorry all, just catching up to all of this now.

So many implications come out of all this, I just don’t know where to start.

on the lighting side alone, I’m trying to figure out… which CMF should be used, when, and for what point of creative decision in the process of lighting a shot?

for acquisition, should cameras encode ultimately for a certain CIE CMF…?? what would be the way to define something like ACES interchange on a spectral basis for an entire series of illuminants and test patches? I spoke with Joseph on this over the phone the other day, and it seems like the only real way to get anywhere with all of this.

for displays… I completely agree that NEW standards must be defined. But… shouldn’t they offer ways to bridge back to old standards…?

1 Like