Digital BW, The Print

Personally, I was always under the impression that "18%" was a convenient
rounding of 18.419%, ie L*=50.  My definition of middle grey would not
deviate much from the way Ansel described it: "a middle gray on a geometric
scale from 'black' to 'white'."  If CIELab bests represents the way we see
then L*=50 is this midpoint.

The reflectance of this middle grey doesn't change with workspace - it is a
defined (reference) grey.  Only the file value that yields this colour
changes.  The more Zonie you are, the more this matters.  The person asking
the questions in the commentary that Ernst quoted has got things a little
muddled or back to front.  When Jon says "Bruce Lindbloom calculated that
gamma 2.46 pegs 18% at L50" I think he has stating things a little back to
front also.  L*=50 is 18.4% reflectance regardless.  What Bruce was doing
was calculating the gamma assumption which would cause an 8-bit file value
of 127/128 to produce a grey of L*=50 which is equivalent to XYZ_Y=18.4, ie
it is one measure of the gamma that best fits CIELab.  Bruce did other
calculations here showing that the best fit gamma was 2.1723 or 2.3243
depending on how you define best fit.

http://www.brucelindbloom.com/index.html?UPLab.html

In practice, I agree with Roy's comments "the ideal thing is to allow anyone
to use whatever they like in the file, profile what your output device does
and use color management to handle the conversion."  This is why I think his
greyscale ICC profiles were such a big step in the right direction and
likely deserve a bit more exploration.  But Zone enthusiasts should note
that this can deviate from the regimen whereby Zone V is a consistent known
quantity from exposure to print. If you perfectly exposed an 18% grey card
and brought it into a Lab workspace it would show L* or K=50 (or 49.5!).
When you printed it, though, the colour of this block of grey can shift
according to the intent used by the colour management system.  If, for
example, one used relcol without black point compensation then the grey
should print as expected.  If one used perceptual or relcol with black point
compensation then the printed shade of grey would be shifted to reflect the
fact that the print can't produce perfect black or white.

Without the use of colour management techniques, the choice of what gamma to
design a print curve to becomes very important.  A printer just gets raw
pixel values (as Roy notes, typically truncated to 8 bit).  The colour
printed for each possible value from 0-255 depends on how the RIP is
calibrated, ie how the curve is constructed. The produced greyscale will
have an implicit gamma or contrast.  If this differs markedly from the space
with which the file is tagged you won't get anything near what you see on
screen.  So, for example, if the greyscale was tuned for a gamma of 1.8 from
ink black to paper white, the overall apparent contrast is still not the
same as you see on screen looking at even a gray gamma 1.8 tagged file
because of the effect of the differing black and white points with a good
display typically have a much better black than matte paper. And the Zone
guys will have to remember that middle grey, 18% reflectance or L*=50, is
not K=50 in their info palette - it's K=61 (but still L*=50 !).  Obviously
things are even worse if you are looking at a file tagged with gray gamma
2.2.  (The gamma at which the display is calibrated at, rather than the
profile embedded with the image, doesn't make a great difference as the
image you see on screen is colour managed by PS.)





> From: "Paul D. DeRocco" <pderocco@...>
> Reply-To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Date: Fri, 17 Jun 2005 01:28:08 -0700
> To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Subject: RE: [Digital BW] Re: Optimal RIP gamma - was how many shades of grey?
> 
>> From: Ernst Dinkla
>> 
>> Quotes:
>> 
>> Can anyone tell me if there is a standard consensus about what Lab
>> Luminosity value a Kodak 18% grey card should yield? I know there is a
>> formula for calculating L* from density. I saw it once but couldn¹t
>> understand it. But seeing it showed me that L* could be directly
>> translated into density. While testing my camera, which I suspect is
>> slipping off the mark, I discovered that the best exposure of a Kodak
>> Q14 greyscale target is not the same exposure that produces L*54 in a
>> 2.2 workspace or L*61 in a 1.8 workspace after conversion through Camera
>> Raw. Now I am wondering if the Kodak 18% grey card is applicable to
>> digital capture. Is there another reflective grey value that is used to
>> calibrate digital cameras and if so, is it standard? Or is something
>> else going on?
> 
> The gamma has no effect on this relation. 18% gray is L=50 (more precisely
> 49.5).
> 
>> Bruce Lindbloom calculated that gamma 2.46 pegs 18% at L50
> 
> A gamma of 2.475 means that a midscale value translates into 18% gray. A
> gamma of 2.44 means that a midscale value equals L=50. So if you accept that
> 18% or L=50 represents the eye's idea of medium gray, a gamma of somewhere
> around 2.45 would be about optimum.
> 
> However, 2.2 is close enough for rock'n'roll. In addition, sRGB has a linear
> segment at the low end, very much like the linear segment in the Lab curve.
> 
> --
> 
> Ciao,               Paul D. DeRocco
> Paul                mailto:pderocco@...
>