[argyllcms] Re: Implementation of setting a black point

Steffen wrote:

Graeme Gill schrieb:


I was just wondering if the intent was to trick the calibration
to have a "dead" zone from zero, and I'm not sure it's really setup
to work that way.
It should work that way. "Black" should begin a little bit earlier so that some headroom is left. 

I don't really agree with you. The idea of calibration I think
is to make the display as responsive as possible, and this means
having a progressive response from zero, so that shadow nuances
are not lost.
I think the gradient doesn't have a consistent white point. There are "stripes" in it. I don't think banding would be the right term, as the gradient is very smooth, but it shows color casts at certain r=g=b values. 


L = #, a=0,b=0 - Whitepoint in [K] (XYZ)
----------------------------------
0    -      7270 (XYZ: 0.13 0.12 0.17)
10    -    5960 (XYZ: 1.76 1.86 1.84)
20    -    6230 (XYZ: 3.69 3.89 4.05)
30    -    6280 (XYZ: 7.24 7.5 8.1)
40    -    6480 (XYZ: 12.74 13.45 14.53)
50    -    6420 (XYZ: 20.77 21.88 23.48)
60    -    6400 (XYZ: 31.74 33.42 35.79)
70    -    6400 (XYZ: 45.38 47.55 51.36)
80    -    6440 (XYZ: 63.45 66.74 72.03)
90    -    6420 (XYZ: 85.07 89.61 96.2)
100  -    6460 (XYZ: 111.32 117.23 126.76)


Turning this into Lab referenced to the white point I get:

   0 -> 0.924640   0.561351 -0.494459
  10 -> 13.148489 -0.147908  1.468520
  20 -> 21.275911 -0.056430  0.805764
  30 -> 30.394392  1.099349  0.031901
  40 -> 40.366253 -0.202670  0.029842
  50 -> 50.291957 -0.031854  0.288489
  60 -> 60.344991  0.016681  0.422452
  70 -> 69.867452  0.619673  0.053280
  80 -> 80.140668  0.162560  0.103816
  90 -> 90.062997 -0.040167  0.438068
 100 -> 100.000000 0.000000  0.000000

So with the exception of the 10% and 30% levels, it's neutral within 1 delta E.
Where would you say the most visible coloring was ?
In neutrals 1 delta E is often discernible, but whether the above result is as
intended depends on what quality level dispcal was set to. In V0.7B8, Medium 
aims for
a maximum error of 0.8 delta E, and High aims for 0.5 delta E.

Graeme Gill.

Other related posts: