LMS and CMF spectral sensitivities?

[Dale]

wikipedia on CIE leaves some details out, so I
have some questions, I did post on the talk page
for
http://en.wikipedia.org/wiki/CIE_1931_color_space
on the subject, COLOR MATCHING FUNCTIONS

I comprehend LongMediumShort color space is
the spectral sensitivity of the cones of the
eye in photopic (not dim) light and that
another space, scotopic rods, enters the picture
in dim lighting

does CIE measure the rods/cones/LMS spectral sensitivity
chemically or by observer experiments?

the way I understand it, color matching functions,
like rbar,gbar,bbar and xbar,ybar,zbar are derived with
observer and lighting experiments

I think rbar,gbar,bbar were designed to produce CIERGB
tristimulus values with the spectral power distributions of colors
perceived

I think, xbar,ybar,zbar were designed for XYZ tristimulus values,
particularly for the point of tristimulus vales with a luminous
function, Y

anyone give an opinion?

[Thomas Richter]

Post by Dale
does CIE measure the rods/cones/LMS spectral sensitivity
chemically or by observer experiments?

CIE defines three curves, and these curves are standardized, not a
measurement procedure. They have only limited correlation to the cone
response curves or the absorption curves of the receptors in the eye.
Let it be as it is, these curves are official, and they are the base of
the XYZ colorspace.

Historically, they have been obtained from color-matching experiments
(with the limited technology available back then), where Y is
(proportional to) the luminance, and then normalizing the curves such
that the area under each curve is one.

Post by Dale
I think rbar,gbar,bbar were designed to produce CIERGB
tristimulus values with the spectral power distributions of colors
perceived.

No. Besides, there is no single RGB space. RGB is a derived colorspace,
based on XYZ given three primaries.

Post by Dale
I think, xbar,ybar,zbar were designed for XYZ tristimulus values,
particularly for the point of tristimulus vales with a luminous
function, Y.

No. X,Y,Z are *defined* by these curves.

Greetings,
Thomas

[Dale]

Post by Thomas Richter

Post by Dale
does CIE measure the rods/cones/LMS spectral sensitivity
chemically or by observer experiments?

CIE defines three curves, and these curves are standardized, not a
measurement procedure. They have only limited correlation to the cone
response curves or the absorption curves of the receptors in the eye.
Let it be as it is, these curves are official, and they are the base of
the XYZ colorspace.
Historically, they have been obtained from color-matching experiments
(with the limited technology available back then), where Y is
(proportional to) the luminance, and then normalizing the curves such
that the area under each curve is one.

Post by Dale
I think rbar,gbar,bbar were designed to produce CIERGB
tristimulus values with the spectral power distributions of colors
perceived.

No. Besides, there is no single RGB space. RGB is a derived colorspace,
based on XYZ given three primaries.

Post by Dale
I think, xbar,ybar,zbar were designed for XYZ tristimulus values,
particularly for the point of tristimulus vales with a luminous
function, Y.

No. X,Y,Z are *defined* by these curves.
Greetings,
Thomas

okay, LMS is still the response of the cones and photopic vision?
scotopic is the rods, dim vision, and there is usually not much
notice of it in the broader fields?

XYZ is "thought" up to be a tri-stimulus space similiar to LMS
with Y being as close to luminousity as possible, and X,Z carrying
the color load?

Xbar,ybar,zbar are the color matching functions used with spectral
power distributions of colors, in color matching experiments, to
get XYZ, and ybar is defined as V the luminousity efficency function
which is close to M in LMS?

XYZ is the "standard observer" (1931,1964?,199sum?) not xbar,ybar,zbar?


the same process in put in place for the CIE RGB space, note your
comment that RGB is more of a implementation, but I think it is
the closest hues that can be associated with LMS

a broader question ...

since CIECAM02 defines LMS as the connection space for appearance
matching, from XYZ, why not get rid of XYZ and use LMS for device
matching too? if you use HSL,HSB,HSV,etc. for color definitions you have a
luminous function, HSV is probably more like what an artist uses to
paint, etc. anyways, ... you can get to device edit controls from HSV
I think ...

an open system would have to be color matching as default, but
I see ICC has some factions and have some pieces/parts of appearance
matching that seems to fit the facts I read

[Thomas Richter]

Post by Dale
okay, LMS is still the response of the cones and photopic vision?

LMS is not XYZ. LMS space is defined by the sensitivity curves of the
cones. Rod vision is (for this purpose) irrelevant.

Post by Dale
XYZ is "thought" up to be a tri-stimulus space similiar to LMS
with Y being as close to luminousity as possible, and X,Z carrying
the color load?

No, once again, XYZ space is not LMS space. The X, Y and Z curves are
"somewhat" arbitrary and not (directly) related to the cone response
curves. Y is lumiance (approximately), X and Z add color information,
but XYZ is not an opponent colorspace (like YCbCr)

Post by Dale
Xbar,ybar,zbar are the color matching functions used with spectral
power distributions of colors, in color matching experiments, to
get XYZ, and ybar is defined as V the luminousity efficency function
which is close to M in LMS?

Well, not really. Y is "observed luminance", and to that, all cones
contribute, more or less. Mostly M, yes, but not only.

Post by Dale
XYZ is the "standard observer" (1931,1964?,199sum?) not xbar,ybar,zbar?

XYZ coordinates are defined by the response curves x,y,z, these define
the "standard observer".

Post by Dale
the same process in put in place for the CIE RGB space, note your
comment that RGB is more of a implementation, but I think it is
the closest hues that can be associated with LMS

There is no "CIE RGB color space". There are multiple RGB color spaces
(sRGB, ITU 601, ITU 709,...).

Post by Dale
since CIECAM02 defines LMS as the connection space for appearance
matching, from XYZ, why not get rid of XYZ and use LMS for device
matching too?

Why would it make any difference? XYZ is standardized, so there are
tables you can use. LMS is not. Depending on the researcher, the
experiment, the observers... LMS space is a little different. But even
if one would standardize LMS space, what difference would it make? LMS
to XYZ is a linear matrix multiplication, thus there isn't much to do,
and there are no advantages.

Besides, XYZ has been defined for various targets in mind, for example
to have always positive coordinates for all physically possible light
sensations.

Greetings,
Thomas

[Dale]

thanks for your time in both responses Thomas

Post by Thomas Richter

Post by Dale
okay, LMS is still the response of the cones and photopic vision?

LMS is not XYZ. LMS space is defined by the sensitivity curves of the
cones. Rod vision is (for this purpose) irrelevant.

Post by Dale
XYZ is "thought" up to be a tri-stimulus space similiar to LMS
with Y being as close to luminousity as possible, and X,Z carrying
the color load?

No, once again, XYZ space is not LMS space. The X, Y and Z curves are
"somewhat" arbitrary and not (directly) related to the cone response
curves. Y is lumiance (approximately), X and Z add color information,
but XYZ is not an opponent colorspace (like YCbCr)

Post by Dale
Xbar,ybar,zbar are the color matching functions used with spectral
power distributions of colors, in color matching experiments, to
get XYZ, and ybar is defined as V the luminousity efficency function
which is close to M in LMS?

Well, not really. Y is "observed luminance", and to that, all cones
contribute, more or less. Mostly M, yes, but not only.

Post by Dale
XYZ is the "standard observer" (1931,1964?,199sum?) not xbar,ybar,zbar?

XYZ coordinates are defined by the response curves x,y,z, these define
the "standard observer".

Post by Dale
the same process in put in place for the CIE RGB space, note your
comment that RGB is more of a implementation, but I think it is
the closest hues that can be associated with LMS

There is no "CIE RGB color space". There are multiple RGB color spaces
(sRGB, ITU 601, ITU 709,...).

Post by Dale
since CIECAM02 defines LMS as the connection space for appearance
matching, from XYZ, why not get rid of XYZ and use LMS for device
matching too?

Why would it make any difference? XYZ is standardized, so there are
tables you can use. LMS is not. Depending on the researcher, the
experiment, the observers... LMS space is a little different. But even
if one would standardize LMS space, what difference would it make? LMS
to XYZ is a linear matrix multiplication, thus there isn't much to do,
and there are no advantages.
Besides, XYZ has been defined for various targets in mind, for example
to have always positive coordinates for all physically possible light
sensations.
Greetings,
Thomas

[Ingo Thies]

Post by Thomas Richter
There is no "CIE RGB color space".

This is not exactly true:

http://en.wikipedia.org/wiki/CIE_1931_color_space#CIE_RGB_color_space

Of course, this is just one among many RGB spaces.

Greetings,

Ingo