I'm trying to get UPLab up and running:
using the LCMS compatible ICC profile version (see page down below).
If I just transform twice, an empty round trip Lab>UPLab>Lab, in float mode, I'd expect the colors not to change as long as there are no challenging gamuts.
However the greens are a little bit washed out and especially on pink color gets pretty desaturated:
(watch the pink with the leaves on it on the circle)
I already tried the Lab2 vs. Lab4 functions and several intents, no help.
My first guess would be that there is some trouble with a LUT point?
Thanks in advance
Tried, and it works quite well to me.
I have converted the JPEG to TIFF in order to avoid compression
artifacts. I'm also using AdobeRGB as source space because the JPEG
has no embedded profile. The procedure is to convert the file
to UPLab and then back to aRGB. Intent is relative colorimetric.
tifficc -t1 -w16 -o AdobeRGB1998.icc -o CIELab_to_UPLab2.icc
tifficc -t1 -e -i CIELab_to_UPLab2.icc -o AdobeRGB1998.icc UPLab2.tif
Please note I'm using 16 bit intermediate storage. Alas, the profile has
some error, if you try the ICC tool (http://color.org/wxProfileDump.zip)
Round Trip 1
Min DeltaE: 0.00
Mean DeltaE: 1.73
Max DeltaE: 26.43
Which is consistent to what you get if you use tiffdiff in the image
tiffdiff -o a.tif adobe.tif StdVsUPLab.tif
Digital counts 99.832% equal. mean 0.428416, min 0, max 13.8564, Std 1.03271
dE Colorimetric 99.8939% equal. mean 0.270681, min 0, max 3.13803, Std
In any case I cannot see big diferences like those in your sample image.
Make sure to use the UPLab profile with correct tag directions
(CIELab_to_UPLab2.icc), as Bruce already states in his website.
The LittleCMS project
Oliver Duis <oduis@...> escribió:
> I'm trying to get UPLab up and running:
> using the LCMS compatible ICC profile version (see page down below).
> If I just transform twice, an empty round trip Lab>UPLab>Lab, in
> float mode, I'd expect the colors not to change as long as there are
> no challenging gamuts.
> However the greens are a little bit washed out and especially on
> pink color gets pretty desaturated:
> (watch the pink with the leaves on it on the circle)
> I already tried the Lab2 vs. Lab4 functions and several intents, no help.
> My first guess would be that there is some trouble with a LUT point?
> Thanks in advance
From: Oliver Duis <oduis@ho...> - 2012-04-10 18:12:36
Good evening Marti,
Thanks for testing! I tried it with Lab in floating point as input/output
(using cmsCreateLab2Profile/4 functions), maybe that's a difference. I also
noticed that the color generally change a little bit, which would explain
the deltas you are measuring. It's just this one color that’s heavily
But to summarize, I understand it's not a bug in my code or LittleCMS, but
the profile that seems to be at least unprecise.
Will check back with Bruce then.