[argyllcms] Re: Host spectral measurements instead of ccmx/ccss?

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 12 Sep 2011 22:50:25 +1000

"Gerhard Fürnkranz" wrote:

> I'm wondering whether the sensor sensitivities of "traditional" instruments 
> could be possibly
> estimated, given the CRT and LCD matrices stored in the instruments and 
> spectral samples for the
> corresponding display types? [Aren't "generic" CRT and LCD samples included 
> in the i1 Display
> pro software?] It's certainly an ill-posed problem which needs to be 
> constrained/regularized,
> e.g. by assuming some a priori knowledge like for instance shape of the 
> sensitivities similar to
> standard observer CMFs, smoothness, positivity, unimodality, etc.

Hi Gerhard,
        the cleverest idea I had in this direction, was to measure the
spectral sensitivities of one or more of each conventional colorimeters,
and save that as a reference along with the corresponding calibration matrix.

Given an instance of a colorimeter of the same type, the approach
would be to compute a correction matrix that transforms the reference
calibration matrix to the instrument matrix, and apply that to the
reference instruments spectral sensitivity curves, to give an approximate
set of curves specific to that instance, and then use that with
the CCSS files to compute a calibration for a particular type of display.

The sticking point of course is being able to measure the spectral
sensitivities of the colorimeters. I certainly don't have the equipment
to do that at this point in time, but in theory it just has to be done
once for each type of colorimeter (although I'm sure some advantage
could be taken of measuring more than one of each type).

Some additional accuracy could probably be gained by determining
a set of CRT and LCD spectra that best explain the reference
colorimeter calibration matrices, and then using that with both
calibration matrices to better estimated the non-reference
colorimeter spectral curves.

With the colorimeter response depending so critically on the interaction
between it's spectral sensitivity curves and the (typically) narrow
peaks in display device spectral output, I'm in some doubt that sensitivity
curves of any merit can be estimated without actually measuring an instrument ?


Other related posts: