[Openicc] L* was: MS on HDR CMS :)

[Chris Murphy]

On Feb 14, 2008, at 4:37 AM, Jan-Peter Homann wrote:
> For a good quality in displaying RGB on a monitor, the RGB  
> workingspace should have the same gamma / gradation as the monitor  
> calibration.

For best quality image reproduction in an 8bpc workflow, you would  
want the TRC of the editing space, the display and the output device  
to be the same (ideally). Each instance of tone mapping involves  
throwing out data. If you tone map to L*, you've tossed more levels  
than had you tone mapped with a gamma 1.8 or 2.2 function. When you  
tone map again to print TRC from L* you lose levels yet again. This  
is not a compelling workflow to me.

Compelling a display to have an L* TRC doesn't seem to be a good idea  
in practice however. Generally we only have 7 or 8 bits to deal with  
in the video card LUT and the curve needed to force a display to L*  
is much more complex than for a simple gamma function.

For 16bpc workflows, that data is always going to be reduced to the  
exact right 8bpc data for the selected TRC for display or output. So  
it doesn't matter if it's linear (capture), L*, or gamma function  
based. So in this instance, it's not a technical problem to use L*, I  
simply don't see it as a benefit. But I do see it as a workflow,  
training, and end user confusion problem.

Now if all you care about is display of images, maybe there's  
something compelling about L* display calibration for a high end  
display that has sufficient precision in the LUT to achieve it, and  
also L* based TRC for the editing space. This is a lot like DICOM.  
It's not a new concept.


> If we look at the standard gamma 1,8 and 2,2 we by converting equal  
> RGB values to Lab, we see that the very dark RGB colors for Gamma  
> 2,2 don´t produce differences and for Gamma 1,8 we have it in the  
> very light colors.

What do you mean it doesn't produce differences? These workflows have  
been around for a rather long time and rather well tested. They do  
have reversibility so I don't understand what, in practice, is  
actually being lost, if anything. Is this with respect to displayed  
data and visual loss somewhere? Is this with respect to actual prints  
and visual loss?


> In Germany several photostudios, photo post production specialists  
> and prepress shops are using eciRGBv2 on monitors calibrated to L*.
> They are very satisfied with the results.

Well I'm sure they're satisfied. I'm sure they were satisfied with  
gamma 2.2 and gamma 1.8 based TRCs as well. The theoretical case so  
far doesn't even make sense to me (yet) let alone the practical case.

Chris Murphy