Re: [Openexr-devel] recent back-and-forth re: Florian's proposal vs. using ICC profiles

[Chris Cox]

At 9:56 AM -0700 8/19/04, Joseph Goldstone wrote:
> (a) Florian's proposal is a scene-referred encoding and although the 
> flexibility to support scene-referred encodings in ICC profiles may 
> be there, it's not the path that is as well trod as that of 
> output-referred encodings.  The ICC needs to write this up, in my 
> humble opinion.

Yes, and I second that motion.



> (c) the ICC was actively reviewing what, if anything, might be 
> needed to support color management for high-accuracy projected 
> imagery coming from a combination of image states (input-referred, 
> output-referred, scene-referred) even before the OpenEXR color 
> management proposal was out there.

Of course, industry involvement and interest might make it go a 
little faster (when there is an expressed need they tend to spend 
less time debating and more time implementing).



> (f) The OpenEXR audience is relatively small compared to that for 
> ICC profiles.  As a consequence, and with all due respect, it may 
> not be as important for OpenEXR that its proposed color management 
> scheme work right the first time; witness the way support for image 
> tiling went into OpenEXR.  To be completely frank, I'd be happy if 
> anything ideas might go into a future ICC revision to support this 
> community's needs be field-tested by means that didn't affect 
> mainline ICC profile users.

Yes, I tend to think of long term solutions:  getting it close to 
right so we do less work overall rather than getting it done now, 
redoing it all later, and being left with bits and pieces of 
incompatible technology on the floor.



> I understand how people like Kevin and Ken have a real, pressing, 
> day-to-day need for image exchange with transportable color.  And as 
> Chris points out, this is what the ICC has been all about since day 
> one.

One thing that could be done is to use the ICC profile as the file 
format for now (since it does have a matrix/TRC form that I believe 
can contain the data proposed by Florian) and not use ICC CMMs for 
the math.  Then, longer term, move the ICC to fully support HDR 
workflows - at which point you could shift to using ICC CMMs for the 
bulk of the work, and adopt more complex profiles and mappings.

That should minimize the duplication of effort, and move toward 
compatibility with other applications and workflows.


Chris