This is something that I do research in for my company. While that’s correct
for the DICOM standard, for other color standards, you need to follow the EOTF
(electro optical transfer function) specified for that standard. (Using Lab
will always be the wrong choice, in fact you’ll notice that it’s not even
called an EOTF.) If you are in a different viewing environment than the
standard (ambient light, peak white of the monitor, etc…), then you need to
adjust the system gamma. Tim Borer of the BBC published research in that area
and it forms the basis of the HLG HDR specification. This is very helpful with
respect to adjusting the system gamma to the peak white. They also did some
work on the ambient light levels, but I think there are issues with that
because it doesn’t match previous research. Specifically, previous
investigations into the optimal system gamma required based on the surround
illumination. For example slide film typically has a system gamma of 1.5
because it is shown in the dark, while prints have a system gamma of 1.0.
Typical video content is usually in the 1.2-1.4 system gamma and assumes a
surround illumination of 5-10% of peak white. I assume that the change in
system gamma is something that is part of the color profiles and used in the
color management routines.
For small changes in the system gamma you can get away with changing the 1D
LUT’s, but this will cause some hue shifts. To do it correctly requires a 3D
LUT or specialized hardware.
Finally, note that the EOTF is not the same as the inverse of the OETF. You
need to find the standard for the EOTF. Almost always (except for DICOM and
HDR), the EOTF is a simple gamma function.
On Jan 22, 2020, at 1:04 PM, <graxx@xxxxxxxxxxxx> <graxx@xxxxxxxxxxxx> wrote:
I stumbled on an article on DICOM, here :
http://umich.edu/~ners580/ners-bioe_481/lectures/pdfs/2008-06-JCI_Fetterly-grayscaleLCD.pdf
<http://umich.edu/~ners580/ners-bioe_481/lectures/pdfs/2008-06-JCI_Fetterly-grayscaleLCD.pdf>
It’s quite impressive, technically. Especially the part regarding “allocating
the available JNDs” to the 8-bit 256 “drive levels”. The way I understand it,
the idea is to make each of the 256 levels result in a “constant” JND.
I’m not about to embark into DICOM-calibrating my monitor but I though it
“confirmed” some of my own intuition on the subject. So, instead of using a
JND, why not use a unit of L* value? So, if my monitor is set for, say, 100
cd/m2 at maximum, and, assume perfect 0 at RGB = 0,0,0 (to simplify
discussions), then the “task” of calibration becomes one of dividing the
available Lightness range (100 units) over the 8-bit available “calibration
slots”, such that 100 divided by 256 is equal to 0.39 L* level? (I’m not
trying to make things more complex than they need to be – believe me) 0.39 is
an odd value (I still have to wrap my head around the implication) but what
if the monitor Max Luminance would be set to 256 cd/m2? Then, the task of
calibration, using that approach, would simply be one of making sure that I
there is exactly *one* Delta L increment between each point of the 8-bit
video LUT/ramp in the video card?
I hate to think out loud like that…
/ Roger
