Monitor calibration and color managed workflow question

On Mon, 19 Dec 2005 05:12:12 -0500, "Ed Ruf (REPLY to E-MAIL IN
SIG!)" wrote:

On Mon, 19 Dec 2005 10:04:53 +0100, in rec.photo.digital Stanislav Meduna
wrote:
....
The Spyder generates an ICC profile, puts it somewhere
into windows directory and registers it with the graphics
card. On startup the ProfileChooser takes the default profile
and loads it into graphics card. That means that I am able
to see accurate colors with applications that are not
color managed - web browser, ...

I thought this was the case as well, but was corrected by Bill Hilton. The
colors on non-managed apps will be better, but not as good as in a color
managed app.

Just checking... What part of the original post needs to be
"corrected"?

I think it is true that a spyder app, or Adobe Gamma, or a third-party
monitor adjustment app, will make a profile and register it with the
OS. And that Adobe Gamma Loader, or a third-party equivalent, will set
graphics card registers at startup so that the "profiled" monitor
shows an sRGB image from any application (or the OS itself) correctly
without any intentional color management. ("Correctly" being to the
limits of whatever process created the profile in the first place, of
course.)

Does a color-managed app fed an sRGB image add any accuracy to this
process? It can't be changing the graphics card registers, or you
would see all the other open windows change color when it took
control. So the color-managed app must be sending unmodified sRGB to
the graphics card, just like a browser would.

Granted, if fed an image in any other color space, the color-managed
app will provide more accurate rendering by intelligently converting
it to the monitor space. Is there any other advantage?

Loren

In article ,
says...
On Mon, 19 Dec 2005 05:12:12 -0500, "Ed Ruf (REPLY to E-MAIL IN
SIG!)" wrote:

Just checking... What part of the original post needs to be
"corrected"?

I think it is true that a spyder app, or Adobe Gamma, or a third-party
monitor adjustment app, will make a profile and register it with the
OS. And that Adobe Gamma Loader, or a third-party equivalent, will set
graphics card registers at startup so that the "profiled" monitor
shows an sRGB image from any application (or the OS itself) correctly
without any intentional color management. ("Correctly" being to the
limits of whatever process created the profile in the first place, of
course.)


No - only a colour management aware app will show the colours correctly.
The RGB values sent to the graphics card need to be adjusted to account
for the monitor profile. Adobe Gamma Loader will only load a
calibration LookUp Table (LUT) to the card, but that is just a starting
point for the colour management process. sRGB is irrelevant (see
below).

Does a color-managed app fed an sRGB image add any accuracy to this
process? It can't be changing the graphics card registers, or you
would see all the other open windows change color when it took
control. So the color-managed app must be sending unmodified sRGB to
the graphics card, just like a browser would.


It converts the colours from the sRGB space to the monitor profile
before sending them to the graphics card. A browser does no conversion
and sends the RGB values "as is"..

Granted, if fed an image in any other color space, the color-managed
app will provide more accurate rendering by intelligently converting
it to the monitor space. Is there any other advantage?


sRGB is just another colour space, in the same way as AdobeRGB or
ProPhoto RGB. There is nothing special about it, and the same arguments
apply for all working spaces.

I suggest you read my other post elsewhere in the thread - I have tried
to explain how monitor profiles work and the difference between
calibration and profiling.

Loren

Stanislav Meduna
The Spyder generates an ICC profile ... (this) means that I
am able to see accurate colors with applications that are not
color managed - web browser, ...

Ed Ruf wrote
I thought this was the case as well, but was corrected by Bill Hilton.
The colors on non-managed apps will be better, but not as good as
in a color managed app.

Loren wrote ...
Just checking... What part of the original post needs to be
"corrected"?

The part about "I am able to see accurate colors with applications that
are not color managed". These apps don't 'convert' the RGB numbers to
get the most accurate colors using the icc profile, while the ICC apps
that recognize monitor profiles do.

Does a color-managed app fed an sRGB image add any accuracy to
this process?

Yes ... as Graeme wrote, all apps benefit from the 'calibration' stage
where you define the gamma and set a white and black point, but then
the profiling software displays color patches on the screen and the
puck measures them and creates a matrix file that enables programs that
recognize monitor profiles to display colors more accurately.

As a theoretical example, most CRT monitors are built such that full or
near full values of the RGB guns give a 9300 K white point but
typically we use 5000 K or 6500 K and to get these warmer colors the
blue and to a lesser extent the green guns are reduced ... to make
something up that I once saw on an older monitor, say 98% R, 80% G, 65%
B to get 5000 K ... this is for the white point, but now it's harder to
get a linear gray, meaning equal RGB values of say 64/64/64 or
150/150/150 are neutral by definition in a 'working space' like sRGB,
but are harder to hit the more uneven the RGB gun values are (like in
the example I gave). So there's often a slight color cast in the
neutrals in non-color managed apps, while this is corrected for in apps
that recognize the profile.

You can turn recognition of the monitor profile on/off in Photoshop
with a keystroke and if I look at a pattern like this one ...
members.aol.com/bhilton665/colors.jpg ... and disable/enable the
monitor profile what I see on my monitor is that most of the squares
look the same but the saturated colors across the top shift noticeably
in the reds, yellows and oranges. I came across this when web photos
from Arizona and Utah's Red Rock country looked different than even
sRGB images in Photoshop as the reds dulled down. Images that weren't
first converted to sRGB looked much duller, but you still see a 2nd
level effect just from the lack of monitor profile recognition.

Bill

Loren writes ...
Granted, if fed an image in any other color space (other than
sRGB), the color-managed app will provide more accurate
rendering by intelligently converting it to the monitor space.

Graeme writes ...
sRGB is just another colour space, in the same way as AdobeRGB or
ProPhoto RGB. There is nothing special about it, and the same
arguments apply for all working spaces.

Actually if you view an sRGB file without the profile (like in a web
browser) there's not much change, but if you view the same file with a
much wider color space in a non-color managed app then the colors can
look REALLY bad ... at least I think that's what Loren is getting at.

Here's an example of a red bird, converted to ProPhoto from the RAW
(this is a very wide space, much wider than AdobeRGB) ... this shows
what the file looks like if I convert to sRGB and then make an untagged
jpeg ...
http://members.aol.com/bhilton665/card_srgb.jpg ... and here's what it
looks like if I just take the ProPhoto tagged file and convert to jpeg
without dumbing it down to sRGB first ... all the saturated colors get
dropped and it looks really really bad ...
http://members.aol.com/bhilton665/card_prophoto.jpg

Granted ProPhoto is a really extreme working space but even with
AdobeRGB you'll see a lot of saturated colors fall out. I think that's
what Loren was getting at.

Bill

Graeme Cogger wrote:

Creating a monitor profile has 2 parts:
1 - Calibration: the display is adjusted (via settings in the graphics
card and maybe the monitor controls) to get things like colour
temperature, gamma etc. correct. All applications see the effect of the
calibration - this is what gets loaded to the graphics card when the
system starts up.
2 - Profiling: the response of the calibrated display is accurately
measured to create a profile of how it displays colour. This profile is
used by profile-aware applications to adjust the colours sent to the
graphics card and monitor. Without these adjustments, the colour
displayed will not be accurate.

Thanks! - that finally makes sense. So the .icc file generated
by the Spyder has actually two parts:

- LUT data tht gets loaded to the card via ProfileChooser
- profile that is used by the PS (and QImage and that's
probably all on my system)

Is the LUT part also standardized somehow? I am asking because
the ProfileChooser refuses to use a profile that it did not
create itself - maybe because it is in some proprietary
format embedded in the vendor-specific part of the ICC?


Thanks to all that responded.

Regards
--
Stano

On 19 Dec 2005 15:28:18 -0800, "Bill Hilton"
wrote:

Stanislav Meduna
The Spyder generates an ICC profile ... (this) means that I
am able to see accurate colors with applications that are not
color managed - web browser, ...

Ed Ruf wrote
I thought this was the case as well, but was corrected by Bill Hilton.
The colors on non-managed apps will be better, but not as good as
in a color managed app.

Loren wrote ...
Just checking... What part of the original post needs to be
"corrected"?

The part about "I am able to see accurate colors with applications that
are not color managed". These apps don't 'convert' the RGB numbers to
get the most accurate colors using the icc profile, while the ICC apps
that recognize monitor profiles do.

Does a color-managed app fed an sRGB image add any accuracy to
this process?

Yes ... as Graeme wrote, all apps benefit from the 'calibration' stage
where you define the gamma and set a white and black point, but then
the profiling software displays color patches on the screen and the
puck measures them and creates a matrix file that enables programs that
recognize monitor profiles to display colors more accurately.

To further check my understanding (or confusion) here...

Am I correct that this "matrix file" is what gets loaded to the lookup
table registers on the graphics card? Or is it a separate color
translation containing subtle differences between what was achieved by
customizing the graphics card lookup table and what the spyder saw?

Maybe I'm behind the state of the art here, but as I understand it,
the graphics card lookup table translates between the three 8-bit
values sent by the application for each pixel, and the three 8-bit
values that get sent on to the monitor to show that pixel "correctly".
If all you have to work with is eight bits, and the graphics card
lookup table has already been set up optimally, what can a color
managed app add to that for an image that is already in the native
color space?

My thinking here is largely based on my experience with an older
notebook where the backlight has gone seriously "warm", to the point
where color errors are obvious. I wasn't able to fix this with Adobe
Gamma, for images inside or outside PS, and concluded Adobe Gamma had
somehow stopped working. I found WiziWYG and it let me tweak the
colors back to some semblance of decency. You can clearly see it load
the graphics registers when its loader runs at startup, and the effect
seems to be the same both inside and outside of PS (though on a
notebook screen maybe the subtleties just aren't visible).

Of course a spyder could do a much better job than I did, and optimize
many more points along each color's correction curve. But wouldn't it
load its conclusions into the graphics registers just like my manual
curves get loaded? And since there are only eight bits per color to
play with, how could a color managed app improve on that? (Unless the
original image needs to be converted to the native color space, of
course.)

I still don't understand what happened to Adobe Gamma, though. It
didn't appear to be able to do anything to help my problem, color
managed app or not. I even tried reloading fresh versions from the CD
and from the web. Does it function differently from WiziWYG?

Loren

In article ,
says...
On 19 Dec 2005 15:28:18 -0800, "Bill Hilton"
wrote:

Loren wrote ...
Just checking... What part of the original post needs to be
"corrected"?

The part about "I am able to see accurate colors with applications that
are not color managed". These apps don't 'convert' the RGB numbers to
get the most accurate colors using the icc profile, while the ICC apps
that recognize monitor profiles do.

Does a color-managed app fed an sRGB image add any accuracy to
this process?

Yes ... as Graeme wrote, all apps benefit from the 'calibration' stage
where you define the gamma and set a white and black point, but then
the profiling software displays color patches on the screen and the
puck measures them and creates a matrix file that enables programs that
recognize monitor profiles to display colors more accurately.

To further check my understanding (or confusion) here...

Am I correct that this "matrix file" is what gets loaded to the lookup
table registers on the graphics card? Or is it a separate color
translation containing subtle differences between what was achieved by
customizing the graphics card lookup table and what the spyder saw?


No, it's separate. The graphics card loads 3 basic curves - one each
for R, G, B. There are 2 main reasons why this is not enough:

- The graphics card knows nothing about the colour space of the original
file, and therefore cannot compensate accordingly. For example, an RGB
value of (200,50,20) in the AdobeRGB space is a different colour to
(200,50,20) in sRGB. If you feed (200,50,20) to the graphics card, it
won't know what colour it needs to display.

- The graphics card Look-Up Table (LUT) has independent curves for R, G
and B. Say you have a monitor that displays greys perfectly, but makes
bright purple, e.g. (200,10,200), appear somewhat blue. The purple
problem can be cured if an R value of 200 is translated to 210 and a B
value of 200 is translated to 190. However, this now means that a pure
grey (200,200,200) will be displayed as (210,200,190) and appear a
little orange. Instead of independent RGB curves, a monitor profile
uses a 3D table and can correct for these issues.

snip

In article , says...
Graeme Cogger wrote:

Creating a monitor profile has 2 parts:
1 - Calibration: the display is adjusted (via settings in the graphics
card and maybe the monitor controls) to get things like colour
temperature, gamma etc. correct. All applications see the effect of the
calibration - this is what gets loaded to the graphics card when the
system starts up.
2 - Profiling: the response of the calibrated display is accurately
measured to create a profile of how it displays colour. This profile is
used by profile-aware applications to adjust the colours sent to the
graphics card and monitor. Without these adjustments, the colour
displayed will not be accurate.

Thanks! - that finally makes sense. So the .icc file generated
by the Spyder has actually two parts:

- LUT data tht gets loaded to the card via ProfileChooser
- profile that is used by the PS (and QImage and that's
probably all on my system)


That's correct

Is the LUT part also standardized somehow? I am asking because
the ProfileChooser refuses to use a profile that it did not
create itself - maybe because it is in some proprietary
format embedded in the vendor-specific part of the ICC?


Unfortunately, the LUT is not standardised. The ICC file format is
basically a set of named tags, and the tags used for the profile itself
are well defined and standardised. There can also be optional tags for
other purposes, which are ignored by colour management engines. Most
profilers use a tag called 'vcgt' to store the LUT, but the way the data
is stored within the tag is different for different apps.


Thanks to all that responded.

Regards

Loren wrote ...

Am I correct that this "matrix file" is what gets loaded to the
lookup table registers on the graphics card?

Graeme answered that well ...

My thinking here is largely based on my experience with an older
notebook where the backlight has gone seriously "warm", to the point
where color errors are obvious. I wasn't able to fix this with Adobe
Gamma ... I still don't understand what happened to Adobe Gamma,
though. It didn't appear to be able to do anything to help my problem,
color managed app or not.

Adobe Gamma does a very poor job with LCDs and laptops ... I thought
this was mentioned in the documentation. Even the original Spyder does
poorly with laptops ... if you're trying to do color-critical work on a
laptop you need to spend $200 or so and get the Eye-One or the Sypder 2
or the Monaco (Optix?) ...

Bill

On Tue, 20 Dec 2005 22:21:25 -0000, Graeme Cogger
wrote:
....
No, it's separate. The graphics card loads 3 basic curves - one each
for R, G, B. There are 2 main reasons why this is not enough:

- The graphics card knows nothing about the colour space of the original
file, and therefore cannot compensate accordingly. For example, an RGB
value of (200,50,20) in the AdobeRGB space is a different colour to
(200,50,20) in sRGB. If you feed (200,50,20) to the graphics card, it
won't know what colour it needs to display.

I think that part was coming clear earlier in this thread.

- The graphics card Look-Up Table (LUT) has independent curves for R, G
and B. Say you have a monitor that displays greys perfectly, but makes
bright purple, e.g. (200,10,200), appear somewhat blue. The purple
problem can be cured if an R value of 200 is translated to 210 and a B
value of 200 is translated to 190. However, this now means that a pure
grey (200,200,200) will be displayed as (210,200,190) and appear a
little orange. Instead of independent RGB curves, a monitor profile
uses a 3D table and can correct for these issues.

Thank you! The 3D profile table as opposed to a 2D graphics card table
is the part I was missing. I'd never noticed that difference in the
color management descriptions I'd read. I suppose I will notice it
everywhere now...

Loren