gimp and ICC color working spaces

[Dale]

gimp has 2 use cases

1) monitor proofing
2) print proofing

yet it only has RGB and CMYK choices for working spaces

I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one

I know there are issues like gamut differences and rendering intents

but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

[Thomas Richter]

Post by Dale
gimp has 2 use cases
1) monitor proofing
2) print proofing
yet it only has RGB and CMYK choices for working spaces
I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one
I know there are issues like gamut differences and rendering intents
but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

The gimp bug report and feature request list is here:

https://bugzilla.gnome.org/buglist.cgi?quicksearch=product%3A%22GIMP%22

[Floyd L. Davidson]

Post by Thomas Richter

Post by Dale
gimp has 2 use cases
1) monitor proofing
2) print proofing
yet it only has RGB and CMYK choices for working spaces
I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one
I know there are issues like gamut differences and rendering intents
but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

https://bugzilla.gnome.org/buglist.cgi?quicksearch=product%3A%22GIMP%22

There is no bug though.

The "RGB" and "CMYK" labels are just that, labels for
two selectable workspaces that the user can easily
switch between.

But the workspace assigned to either of those two labels
can be
1) virtually anything the system administrator
has installed and made available, or
2) anything the user chooses to install privately.

Any user that wants to use a different workspace merely
needs to download whatever it is and install it. In
GIMP, select the Edit->Preferences->Color Management
menu. Both the "RGB" and "CMYK" profile options have a
selection box, and the last option in either box is to
"Select color profile from disk". Once a profile has
been selected it will then show up as one of the
available options and choosing it will not require
locating it again.

Sources for such profiles include LCMS, Argyll, and others.

[Dale]

Post by Floyd L. Davidson

Post by Thomas Richter

Post by Dale
gimp has 2 use cases
1) monitor proofing
2) print proofing
yet it only has RGB and CMYK choices for working spaces
I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one
I know there are issues like gamut differences and rendering intents
but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

https://bugzilla.gnome.org/buglist.cgi?quicksearch=product%3A%22GIMP%22

There is no bug though.
The "RGB" and "CMYK" labels are just that, labels for
two selectable workspaces that the user can easily
switch between.
But the workspace assigned to either of those two labels
can be
1) virtually anything the system administrator
has installed and made available, or
2) anything the user chooses to install privately.
Any user that wants to use a different workspace merely
needs to download whatever it is and install it. In
GIMP, select the Edit->Preferences->Color Management
menu. Both the "RGB" and "CMYK" profile options have a
selection box, and the last option in either box is to
"Select color profile from disk". Once a profile has
been selected it will then show up as one of the
available options and choosing it will not require
locating it again.
Sources for such profiles include LCMS, Argyll, and others.

you are right, my post was not a bug report, I was trying to prompt a
feature discussion

RGB aand CMYK are sensitometric and densitrometric spaces respectively,
and are linear withh with respect to light

I was suggesting a colorimetric working space like CIELAB or CIELUV
because delta E* is linear with just noticeable differences of the eye
and can be measured with a colorimeter or spectrophotometer or
spectroradiometer

[Dale]

Post by Dale

Post by Floyd L. Davidson

Post by Thomas Richter

Post by Dale
gimp has 2 use cases
1) monitor proofing
2) print proofing
yet it only has RGB and CMYK choices for working spaces
I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one
I know there are issues like gamut differences and rendering intents
but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

https://bugzilla.gnome.org/buglist.cgi?quicksearch=product%3A%22GIMP%22

There is no bug though.
The "RGB" and "CMYK" labels are just that, labels for
two selectable workspaces that the user can easily
switch between.
But the workspace assigned to either of those two labels
can be
1) virtually anything the system administrator
has installed and made available, or
2) anything the user chooses to install privately.
Any user that wants to use a different workspace merely
needs to download whatever it is and install it. In
GIMP, select the Edit->Preferences->Color Management
menu. Both the "RGB" and "CMYK" profile options have a
selection box, and the last option in either box is to
"Select color profile from disk". Once a profile has
been selected it will then show up as one of the
available options and choosing it will not require
locating it again.
Sources for such profiles include LCMS, Argyll, and others.

you are right, my post was not a bug report, I was trying to prompt a
feature discussion
RGB aand CMYK are sensitometric and densitrometric spaces respectively,
and are linear withh with respect to light
I was suggesting a colorimetric working space like CIELAB or CIELUV
because delta E* is linear with just noticeable differences of the eye
and can be measured with a colorimeter or spectrophotometer or
spectroradiometer

CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs

the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

[Thomas Richter]

Post by Dale
CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs
the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

First, this is a *PROFILE* connection space. It is unrelated to
*PERCEPTUAL* rendering. Once more!

Second, it is *UNRELATED* to ideal printing. The PCS has nothing to do
with printing, rendering intent, the material you print on or its
properties.

Third, the PCS doesn't have a "gamut". XYZ is able to represent all
colors, if you like to.

[Dale]

Post by Thomas Richter

Post by Dale
CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs
the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

First, this is a *PROFILE* connection space. It is unrelated to
*PERCEPTUAL* rendering. Once more!
Second, it is *UNRELATED* to ideal printing. The PCS has nothing to do
with printing, rendering intent, the material you print on or its
properties.
Third, the PCS doesn't have a "gamut". XYZ is able to represent all
colors, if you like to.

you are confusing the PCS with the CMM, the PCS does have rendering
choices that you make on input and output profiles

[Thomas Richter]

Post by Dale

Post by Thomas Richter

Post by Dale
CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs
the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

First, this is a *PROFILE* connection space. It is unrelated to
*PERCEPTUAL* rendering. Once more!
Second, it is *UNRELATED* to ideal printing. The PCS has nothing to do
with printing, rendering intent, the material you print on or its
properties.
Third, the PCS doesn't have a "gamut". XYZ is able to represent all
colors, if you like to.

you are confusing the PCS with the CMM, the PCS does have rendering
choices that you make on input and output profiles

*Sigh*, no I'm not confusing this. I've already written CMMs, if you
care. The PCS does not have rendering choices in the same sense that the
R^3 does not have vector choices. It is the CMM that can implement
various rendering intents.

[Martin Leese]

Dale is just a troll. Please stop feeding
the trolls.

[Dale]

Post by Martin Leese
Dale is just a troll. Please stop feeding
the trolls.

go to alt.usenet.kooks and recommend me for a kook award

I haven't had one in years

[Dale]

Post by Dale

Post by Martin Leese
Dale is just a troll. Please stop feeding
the trolls.

go to alt.usenet.kooks and recommend me for a kook award
I haven't had one in years

there IS a cabal

[Dale]

Post by Thomas Richter

Post by Dale

Post by Thomas Richter

Post by Dale
CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs
the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

First, this is a *PROFILE* connection space. It is unrelated to
*PERCEPTUAL* rendering. Once more!
Second, it is *UNRELATED* to ideal printing. The PCS has nothing to do
with printing, rendering intent, the material you print on or its
properties.
Third, the PCS doesn't have a "gamut". XYZ is able to represent all
colors, if you like to.

you are confusing the PCS with the CMM, the PCS does have rendering
choices that you make on input and output profiles

*Sigh*, no I'm not confusing this. I've already written CMMs, if you
care. The PCS does not have rendering choices in the same sense that the
R^3 does not have vector choices. It is the CMM that can implement
various rendering intents.

we kind of got off my original post, can gimp or do other osftwares have
colorimetric working spaces like CIELAB, CIEXYZ, etc., as opposed to
spaces like RGB and CMYK?

[Dale]

Post by Thomas Richter

Post by Dale

Post by Thomas Richter

Post by Dale
CIECAM would be better, of course you have to pick a standard CAM
(appearance) for the profile connection space to convert between CAMs
the current ICC perceptual connection space uses an ideal print for the
CAM, this is not ideal for transparent and translucent display materials
or additive systems, they have a larger gamut than a print and you have
to make up information when you render from the standard CAM (the print)
to these gamuts, maybe you can have choices for the standard CAM

First, this is a *PROFILE* connection space. It is unrelated to
*PERCEPTUAL* rendering. Once more!
Second, it is *UNRELATED* to ideal printing. The PCS has nothing to do
with printing, rendering intent, the material you print on or its
properties.
Third, the PCS doesn't have a "gamut". XYZ is able to represent all
colors, if you like to.

you are confusing the PCS with the CMM, the PCS does have rendering
choices that you make on input and output profiles

*Sigh*, no I'm not confusing this. I've already written CMMs, if you
care. The PCS does not have rendering choices in the same sense that the
R^3 does not have vector choices. It is the CMM that can implement
various rendering intents.

a profile can be populated with 4 intents, perceptual relative to a
print, maintain saturation, colorimetric and absolute colorimetric

the PCS is instantiated to represent the intents and conversion spaces
chosen or specified

[Thomas Richter]

Post by Dale

Post by Thomas Richter
*Sigh*, no I'm not confusing this. I've already written CMMs, if you
care. The PCS does not have rendering choices in the same sense that the
R^3 does not have vector choices. It is the CMM that can implement
various rendering intents.

a profile can be populated with 4 intents, perceptual relative to a
print, maintain saturation, colorimetric and absolute colorimetric
the PCS is instantiated to represent the intents and conversion spaces
chosen or specified

No, it's not, for the last time! An *ICC profile* includes a rendering
intent. It *also* defines a profile connection space, but the profile
connection space has no rendering intent.

How often do I need to explain you that the PCS is just a common
coordinate space to express colors both of the input and the display (or
output) device?

The ICC profile contains enough information to convert from the device
color space to the PCS and back. It defines the type of the PCS, it also
defines the rendering intent which, again, defines the peculiarities in
how to interpret the coordinate transformation (or how or where to find
it). But the PCS is nothing but the coordinate frame in which colors are
expressed. It doesn't have an intent. If coordinate transformations
could be done in infinite precision, and if we wouldn't care about
practical implementation limits, the PCS would not even matter. We could
pick XYZ for everything and express coordinates always in XYZ.

The CMM is the piece of software that implements the coordinate
transformation to and from the PCS. It implements the rendering, and by
that also defines how to realize the rendering intent - it has some freedom.

But why the heck don't you just download the specs from color.org and
read them yourselves?

[Dale]

Post by Thomas Richter
The ICC profile contains enough information to convert from the device
color space to the PCS and back

the profile intent choices define the PCS and instantiate it in the CMM

[Thomas Richter]

Post by Dale

Post by Thomas Richter
The ICC profile contains enough information to convert from the device
color space to the PCS and back

the profile intent choices define the PCS and instantiate it in the CMM

No, it doesn't.

[Dale]

Post by Thomas Richter

Post by Dale

Post by Thomas Richter
The ICC profile contains enough information to convert from the device
color space to the PCS and back

the profile intent choices define the PCS and instantiate it in the CMM

No, it doesn't.

then how does the CMM connect intent choice on both the input and output

[Thomas Richter]

Post by Dale

Post by Thomas Richter

Post by Dale

Post by Thomas Richter
The ICC profile contains enough information to convert from the device
color space to the PCS and back

the profile intent choices define the PCS and instantiate it in the CMM

No, it doesn't.

then how does the CMM connect intent choice on both the input and output

Why don't you just download the ICC specs from color.org and study it?

The ICC profile defines how to map the device colorspace to the PCS by
including tables and/or parameters for the transformation.

The PCS is constant, but depending on the ICC profile, more than one
table/parameters exists how to implement the mapping. The CMM checks the
rendering intent, and then picks table A, B or C that transforms the
device colors to the PCS coordinates. The PCS stays always the same,
typically XYZ. Despite this explicit dependency, the CMM may also
(implicitly) make rendering-intent specific algorithmic choices, for
example how to handle out-of-gammut colors or how to implement an
adaption on the illumination source.

Greetings,
Thomas

[Dale]

Post by Thomas Richter

Post by Dale

Post by Thomas Richter

Post by Dale

Post by Thomas Richter
The ICC profile contains enough information to convert from the device
color space to the PCS and back

the profile intent choices define the PCS and instantiate it in the CMM

No, it doesn't.

then how does the CMM connect intent choice on both the input and output

Why don't you just download the ICC specs from color.org and study it?
The ICC profile defines how to map the device colorspace to the PCS by
including tables and/or parameters for the transformation.
The PCS is constant, but depending on the ICC profile, more than one
table/parameters exists how to implement the mapping. The CMM checks the
rendering intent, and then picks table A, B or C that transforms the
device colors to the PCS coordinates. The PCS stays always the same,
typically XYZ. Despite this explicit dependency, the CMM may also
(implicitly) make rendering-intent specific algorithmic choices, for
example how to handle out-of-gammut colors or how to implement an
adaption on the illumination source.
Greetings,
Thomas

I think I got it now, Thanks

[Dale]

Post by Thomas Richter

Post by Dale
gimp has 2 use cases
1) monitor proofing
2) print proofing
yet it only has RGB and CMYK choices for working spaces
I don't know if Photoshop or other software has a CIE colorimetric
choice for working space, but in order to measure and track color it
would be a lot easier to have one
I know there are issues like gamut differences and rendering intents
but if you want to measure the colorimetry of the monitor or proof and
match it to the print, you need to start with CIE, not RGB or CMYK,
ideally CIECAM but ICC has not gone there yet as far as I know

https://bugzilla.gnome.org/buglist.cgi?quicksearch=product%3A%22GIMP%22

did you say before that gimp's CMM was 8bit?

8bit encompasses the visual range of colors, but processing, to meet
Nyquist would be 16bit or higher

I used to make color profiles and know that few have precision of 1 delta E*

I also know there is little difference between good enough color and
perfect color, and that some things are better off left to editing than
specification

[Thomas Richter]

Post by Dale
did you say before that gimp's CMM was 8bit?

Yes, it's to my very knowledge .

Post by Dale
8bit encompasses the visual range of colors, but processing, to meet
Nyquist would be 16bit or higher

That's not a frequency, so it's not related to *Nyquist*. But yes,
depending on the steep of the tone mapping curves, you might get banding
(quantization artifacts) due to the limitation of the resolution.