[argyllcms] Re: Shadow detail problem
- From: "Dave Wagner" <dave.wagner@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 1 Feb 2008 12:21:24 -0600
Thanks Graeme for the quick reply.
On Feb 1, 2008 5:29 AM, Graeme Gill wrote:
> Dave Wagner wrote:
>
>
> > However, in the resulting profiles the shadow area is pushed toward
> > black. The first viewable black level is around 21 (RGB). This is
> > not at all like the results on my CRT at home.
>
> How are you determining this ? Are you speaking about the
> vcgt tag in the profile (the calibration), or the
> profile ? The latter doesn't in itself change color,
> it only has an effect when linked to another device
> profile using a CMM (typically the input profile).
Here is what I am doing, I create a black (rgb 0,0,0) image in gimp
and then select a rectangle and bump it one rgb level at a time until
I can see a difference between it and the background. With just the
calibration loaded (.cal file or vcgt tag, it is visibly different at
3,3,3 or 4,4,4 (of 255 for 8-bit in gimp). With the CMM loaded, it
isn't visibly different until 21,21,21.
> > So, what am I doing wrong? Or, is it correct for it to flatten out the
> > shadows?
>
> You need to investigate what your system is doing with the profile,
> and what you expect of it. Assuming for the moment that the vcgt tag
> calibration is being loaded into your system somehow (using
> dispwin etc), then if (for instance) your CMM (Whatever Gimp
> is using in this case, lcms ?) is taking a grey wedge defined in RGB,
> interpreting that in some idealized color space (such as sRGB), and
> then doing a relative colorimetric translation of those colors
> into your LCD display colorspace, then yes, the shadows will
> drop out because the LCD display can't display an ideal 0 black
> that sRGB colorspace is.
> It won't display anything until the sRGB L* value reaches about 7.
>
> If you want some other sort of behaviour (ie. that the black
> level of the idealized input space be smoothly gamut mapped),
> then this has to be arranged and asked for somehow, if Gimp
> doesn't do this by default.
>
> One approach might be to use some sort of CMM link time
> grey axis gamut mapping, such as "black point compensation".
> (does lcms and Gimp provide this as an option for the
> screen rendering ?)
>
> Another might be to setup the display profile to gamut map
> from sRGB using the ICC CLUT gamut mapping mechanism.
> (Note that shaper/matrix profiles don't support this
> type of gamut mapping, they only support colorimetric
> rendering natively). If your source colorspace was
> sRGB, then you'd have to specify this as the source
> colorspace when making the profile (Argyll profile -S option),
> and then select perceptual rendering in your CMM (assuming Gimp
> has such control over it's display color rendering).
>
I have tried all 4 rendering intents with no noticeable change in
shadow detail. I have gimp setup to use sRGB colorspace (because gimp
is 8-bit per channel and because I usually print on Fuji lightjets
which have a gamut that is more similar to sRGB than the larger color
spaces.) I have the same results with profiles built with sRGB as a
source gamut and without.
However, if I build a shaper profile for the LCD panel instead of a
LUT, it does not do this. It has a similar response to the monitor
with calibration only loaded (shadow is distinguishable from black at
rgb 4,4,4). This shaper profile is built from the same 1500 patch
.ti3 file.
I was under the impression that a LUT was superior for an LCD panel.
This is an S-IPS panel, so maybe it's color response is well behaved
enough to just use a shaper profile...
Would generating a .ti1 file with gray axis patches in addition to the
normal iterative patches help the accuracy of the LUT profile in the
shadow region?
I guess my concern is that either the contrast displayed in the shaper
profile OR the contrast displayed in the LUT profile is closer to
"accurate" for what I'll get when I print on an fuji lightjet. I
suppose the best test is to take a printed picture and compare :)
Also, this is not limited to Gimp, it gives the same results through
ufraw (which also uses LCMS for color management). The shaper profile
gives more shadow detail there too. I do not believe that the
cinepaint glasgow for windows has the CMS it does on linux, or I
would try it too. Too bad I can't run linux at the office too :)
Thanks again for your time and your help.
Other related posts:
