Re: Argyll CMS and tiffgamut
Re: Argyll CMS and tiffgamut
- Subject: Re: Argyll CMS and tiffgamut
- From: "dpascale" <email@hidden>
- Date: Thu, 21 Feb 2008 22:41:22 -0500
Hello Graeme,
I'm unable to duplicate your results. If I compare (say) sRGB
and AdobeRGB1998 profiles on the same RGB .tif file, I see a noticeable
difference in both the gamut hulls computed, and the resultant reported
volume.
This is not the problem I have; I obtain the same results as you when I make
the test you just described since my data's gamut exceeds the size of the
profile I select. My question is when using the same profile and two RGB
images based on different RGB spaces.
The gamut is computed from the .tif file, interpreting it's device color
space
into L*a*b* or CIECAM02 Jab space using the given profile with the given
intent.
[Note that currently none of the Argyll utilities pay attention to
embedded profiles in .tiff files, since I haven't figured out
how to best approach it from a UI perspective, and in general
it's better to specify what's going on during diagnosis, rather
than things happening invisibly.]
This explains what I see.
So, when it sees RGB, what RGB space is assumed?
Note: I tried using a L*a*b* TIF file but this is not supported.
With the -v (verbose option), the total
volume of the constructed surface is reported.
Tried again but I do not see it, neither in the command window or in the
.gam file. I do see two warnings but the .gam file is made nonetheless,
which shows it has done its work. I also obtain VRML images that I can
compare in a viewer, so that I can see if and where the differences are.
If you are using some sort of utility that is color aware,
and converts the raster file from one RGB colorspace to
another by maintaining the CIE values, then of course
there will be no difference in the resulting
gamut or volume, since the color hasn't changed,
just the colorspace used to represent it. I'm not sure
if that is in fact what happened in your testing or not.
No, I did not convert to a profile. I assigned two different RGB spaces to
the same data, and saved each file with its assigned profile.
Thanks for your response,
Danny
email@hidden
----- Original Message -----
From: "Graeme Gill" <email@hidden>
To: <email@hidden>
Sent: Thursday, February 21, 2008 9:58 PM
Subject: Re: Argyll CMS and tiffgamut
dpascale wrote:
It does not compute the gamut volume as a single number, but gives a
gamut extent in the form of L*a*b* coordinates which enclose a volume
(look for a .gam file). It can also generate a VRML 3D image (.wrl) of
the same data.
The gamut is computed from the .tif file, interpreting it's device color
space
into L*a*b* or CIECAM02 Jab space using the given profile with the given
intent.
Since a raster file is a collection of point samples, an enclosing surface
is computed from the points, the surface being formed to meet certain
criteria
in regard to shape (the surface is radial from the 50,0,0 point), and
has a limit imposed on its maximum concavity. This means that the gamut
surface is an approximation. With the -v (verbose option), the total
volume of the constructed surface is reported.
The primary purpose of this utility is not to estimate the gamut
volume of an image, but to construct a representative source
gamut for image specific gamut mapping.
The way I understand tiffgamut is that it wraps an icc profile around the
image gamut, with the maximum gamut being the profile itself. As a
consequence, all other image data, out of the profile gamut, is clipped.
By definition the profile limits the gamut, since it determines
the conversion of device values to CIE values, and it and
the raster file format limit the device value range.
Note that some extreme RGB profiles have primaries that
exceed the ICC L*a*b* representational range (a*b* -128 .. 128),
and that this limit may clip the result.
I made the following test:
I saved two TIF files containing the same RGB coordinates, but one file
assigned with the sRGB profile and the other with ProPhoto.
Since the ProPhoto file describes a larger gamut than the sRGB one, I
expected a larger enclosed gamut. However, the gamut files are exactly
the same. I used Relative Colorimetric (-ir flag) or Abs-Col (-ia)
without significant differences. I asked the question to the utility
author and am waiting for a response.
I'm unable to duplicate your results. If I compare (say) sRGB
and AdobeRGB1998 profiles on the same RGB .tif file, I see a noticeable
difference in both the gamut hulls computed, and the resultant reported
volume.
[Note that currently none of the Argyll utilities pay attention to
embedded profiles in .tiff files, since I haven't figured out
how to best approach it from a UI perspective, and in general
it's better to specify what's going on during diagnosis, rather
than things happening invisibly.]
If I compare -ia and -ir gamuts using viewgam, there is
a quite noticeable difference. Note I used the original HP/Microsoft
sRGB profile for this test, and that many other ICC RGB profiles
do not record their actual absolute white point in the white
point tag, and hence give relative colorimetric rendering
even when absolute colorimetric is selected.
Note also that the sRGB ICC V2 profiles available on the ICC
website are (in my opinion) not compliant with the published
ICC V2.4 specification since they manipulate the black
point in particular ways, and have been known to give strange
results in some situations. They are examples also of
profiles that record their white point in such a manner
that absolute intent rendering behaves exactly the same
as relative intent, rather than returning instrument readings.
If you are using some sort of utility that is color aware,
and converts the raster file from one RGB colorspace to
another by maintaining the CIE values, then of course
there will be no difference in the resulting
gamut or volume, since the color hasn't changed,
just the colorspace used to represent it. I'm not sure
if that is in fact what happened in your testing or not.
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden