[argyllcms] Re: colormouse vs. digital camera

  • From: Gerhard Fuernkranz <nospam456@xxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 19 Oct 2005 23:13:33 +0200

Stephan Bourgeois schrieb:

2. why is there more color cast with the colorimeter than the digital camera method? Hypothesis: the colormouse is more accurate and my screen is wrong. But then I shouldn't have a color cast when I print a gray scale.

Just a couple of thougts:

* Did you give the prints enough time to dry?
 (depending on the used paper and inks, the colors
 can still change significantly within 24 hours
 after printing)

* Under which light source do you judge the prints?

* Can you also MEASURE the presence of the color cast with
 your colorimeter, or can you only SEE it? If you can also
 measure a significant error, then the profile does not
 describe your printer's characteristics accurately enough.
 If you measure only an insignificant deviation (i.e. if
 the profile is accurate), then either your viewing
 illuminant or the instrument accuracy are likely the
 reason for the color cast you can see.

 Example how to check:

 # find printer RGB for a middle gray patch, L*=50
 $ xicclu -ir -fif -pl myprinter.icc
 50 0 0
 50.000000 0.000000 0.000000 [Lab] -> Lut
   -> 0.326174 0.428683 0.439484 [RGB]

 # Compute absolute L*a*b* for these RGB values.
 # That's what your colorimeter should read, if
 # you would print this patch and measure it.
 $ xicclu -ia -ff -pl myprinter.icc
 0.326174 0.428683 0.439484
 0.326174 0.428683 0.439484 [RGB] -> Lut
   -> 44.778623 0.026997 -3.770875 [Lab]

 # multiply RGB numbers by 255
 # (convert to 0..255 range)
 $ bc -l

 ==> print a patch with RGB = [83,109,112] and measure
 it with the colorimeter (after drying). If the profile
 describes your printer accurately, then your colorimeter
 should read L*a*b* = [44.778623, 0.026997, -3.770875]

* You said, that you only used 220 patches to create the
 profile. This may not be enough to achieve an accurate
 profile with a good, neutral gray axis. If I resample
 (fakeread) my printer profile with only 220 full-spread
 samples, and create a profile from these 220 samples,
 then then the gray axis of the resampled profile also
 deviates noticably from my original profile.

 If the low number of patches is the reason for the
 inaccuracy, then you could try to augment your 220
 patches with e.g. 100 additional patches along the gray
 axis, and create a new profile from all 320 patches.
 You can use "xicclu -ir -fif -pl ..." (see above) to find
 the RGB numbers for the additional gray axis patches.

Btw, Graeme,

IMHO it would be nice, if "targen -d3 -c <profile> -g <steps> ..." would not generate patches with R=G=B, but if it would instead create a ramp of true neutral patches (i.e. a ramp of patches with a* = b* = 0 in the PCS, according to the relcol intent of the given preliminary profile).


Other related posts: