Re: Photoshop and LUT Profiles
Re: Photoshop and LUT Profiles
- Subject: Re: Photoshop and LUT Profiles
- From: Peter Karp <email@hidden>
- Date: Mon, 10 Jan 2005 16:44:05 +0100
Dear Marco and Roger,
I'm working for Quato and understand your questions. Normally I read
and post in the mailing list as a private person and I strongly try to
avoid to post "marketing hype" or "advertisements". I'm very lucky
that this list is an open forum to communicate and learn about color
related things and I try to contribute when I think my answer can be
helpfull or interesting for others.
Let me try to answer yours and Rogers posting in one e-mail:
> While reading the "Calibration White Paper" PDF for the iColor Proof IP
> software that drives the Quato LCD displays, I came across this statement
> (on page 4 of the document, available at
> <http://www.quato.de/english/products/whitepaper_quato_e.pdf>):
>> "As the world¹s first dedicated display calibration software, iColor Proof IP
>> allows you to save the profile in two different ways without the need of a
>> recalibration. One can choose between a matrix- and a LUT-profile. As Adobe
>> Photoshop is not able to work correctly with LUT-profiles it¹s not recommended
>> to use this kind of profile in such a workflow. Instead a matrix-profile will
>> do the job."
> Now, I am no wiz when it comes to the inner arcana of color profiles, but
> doesn't this sound a bit off?
Yes, it does sound off, but it's correct. You can choose to work with
a LUT based monitor profile with Photoshop. There are advantages and
disadvantages with a LUT profile. In general Photoshop can work with
LUT-monitor profiles, but I'll explain what's the problem in PS.
Photoshop reads LUT based profiles correct and -- in contrast to some
other applications -- can use a LUT monitor profile and not only a
matrix-shaper profile. The statement that photoshop can not work
correctly with LUT monitor profiles doesn't mean that LUT profiles are
not supported, but that there are problems or quirks in Photoshop.
When you open a tagged file in PS and convert the picture to another
profile and save this picture in the new color space you'll get good
results. This is true for matrix and LUT monitor profiles. Of course
you'll seldom find a case where it makes sense to convert a file to
the color space of a specific display, but nonetheless this would work
fine.
But the usual usage of a monitor profile of Photoshop is to use the
monitor profile to calculate the colors to display them in such a
way that the meaning of the color (which is given in a ICC-profile
tagged file) is preserved. But using a LUT monitor profile will lead
to errors when displaying a picture in PS.
Those errors will often not be visible, but especially in the shadow
area of a picture and/or smooth gradients the errors will show up.
You'll see banding in areas which should look smooth, because the file
contains a smooth gradient. But with a LUT profile you'll see some
banding very clearly and you might think that the monitor calibration
isn't good and you can not easily deceide wether the banding occurs
a) because the data in the file simply has this banding and the data
is not smooth
b) or because the monitor was calibrated "wrong" (not enough steps
like in some 6-Bit panels of a laptop LCD)
c) or because Photoshop introduces the banding when displaying the
picture.
You can try this for yourself:
Open a true non-dithered grayramp in PS. I often use a grayramp where
I have some pixels for each gray tone. For example I created a file
with 3 pixels for each tone that will look like this:
0 0 0 1 1 1 2 2 2 3 3 3 4 4 4 .... 254 254 254 255 255 255
... the same for further rows ....
Or I have other files with 1 or 5 or 6 pixels for each gray color.
When you look at such an artificial gray ramp you're very likely to see
some banding on most displays. But you have to be aware that this
banding can have the three above mentioned causes. When you view such
a file in a program which doesn't account for the monitor profile
(such as nearly all simple picture and viewing programs for Windows
and most for Mac too) the banding is either in the file itself or
introduced because of a bad monitor calibration (either in the
hardware of the display or via the calibration curves which are often
loaded onto the video graphics card) [1].
Provided a good monitor calibration and smooth gray ramp data you'll
see that Photoshop introduces the banding for the display output when
you use a LUT monitor profile.
To verify this problem I have created a 1 by 256 pixel file with one
pixel for each possible level in an 8-bit file:
0 1 2 3 4 .... 254 255
I tagged this file with AdobeRGB. When you use a matrix _or_ a LUT
profile for the monitor and you convert the file to the color space of
the monitor you'll end up with exactly the same file:
0 1 2 ... 255
But when you view this file in PS and measure the values which are
sent to the display [2] you'll get very different results. With a
matrix profile you'll loose some levels in the deep shadows. Midtones
and highlights are not altered:
0 2 2 4 4 6 6 8 8 10 10 11 11 13 13 15 16 17 18 ... 254 255
You will still have 249 tone levels from the original 256 levels.
Now comparing to the file when it's displayed with a LUT profile:
14 14 14 14 16 16 16 16 16 16 18 18 18 18 18 20 20 20 22 22 ... 255
This time you end up with only 214 levels! You'll also notice that you
loose black and that up to the midtones some levels are dropped.
Looking at this gray ramp will show _very_ clearly visible banding and
a different tonal response in the shadows compared to the matrix
profile.
For this test the matrix and the LUT profile are both of good quality
and you see that the profile type (LUT vs. matrix) is not responsible
for the problem, because Photoshop itself handles the LUT profile
correct when converting the files. You can double check this using
other CMM's. They also give similar results (sometimes they loose some
levels, but not like PS when displaying files "through" a LUT
profile). Photoshop should display the files the same way like when
converting the files, but this is _not_ true.
I understand that PS might have to make a compromise in terms of
quality and time/speed to display/update a picture, but the problems
with LUT monitor profiles are very serious IMO and the difference to a
matrix profile is much too big. Binuscan Photoretouch Pro for the Mac
doesn't show this problem with LUT monitor profiles. That's a sample
that it can be done better and that the LUT profile itself is not the
cause of those problems.
Those PS problems -- not correctly displaying a picture when a LUT
monitor profile is used -- are clearly visible in such an artificial
test environment, but will also show up in real pictures. Of course
color pictures are also affected by this problem.
One last example for an arbitrary blue color (tagged with AdobeRGB) in
Photoshop:
Matrix profile conversion: 26 59 150 (RGB values)
LUT profile conversion: 20 54 150
You see that the LUT and matrix profile will end up in different
values. This is not surprising, because a LUT profile has the chance
to describe the color space more exact and you'll end up with slightly
different (when the programmer has done a good job: more exact)
values.
But when viewing the AdobeRGB file in PS you should of course get the
same values like when converting the file. The sample color will be
displayed like:
Matrix profile display: 26 59 150
LUT profile display: 31 55 150
The RGB values noted here are the values which are sent to the display
and _not_ the values of the file. It's important to understand the
difference.
Again you'll notice that the LUT display and the LUT conversion of the
file differ. In addition the difference to the matrix profile goes
into two different directions for conversion/display. That's a
problem. I double checked the conversion sample color with the CMM
from Argyll and found argyll to give the very same values like
Photoshop's conversion values.
> And (b) what is this about Photoshop not being "able to work
> correctly with LUT-profiles"? Is this a load of bull or is there any
> truth to it?
I hope it's now understandable why it's stated that Photoshop can not
work correct with LUT monitor profiles.
> From what I know so far, (a) LUT-based monitor
> profiles are supposed to be BETTER than matrix-based ones, given the
> latter's inherently imprecise rendition of the display's behavior.
Yes, that's true often. A LUT profile _can_ be more accurate then a
matrix profile (it can also be worse). In the case of the Quato iColor
software LUT profiles are more accurate and we will advise in general
to use them when this is possible.
But I personally advise to test both profile types in your specific
environment with the applications you're using and to choose the
profile type which works best for you.
I myself choose matrix profiles normally. I do so not because they are
better, but because they are less prone to result in errors due bugs
in applications. Or some applications simply can not use LUT monitor
profiles. One example is/was the Kodak camera software. (We reported
this to Kodak and I'm not sure if it's already fixed). In addition a
matrix profile can describe an accuratly hardware calibrated display
like the the Intelli Proof very good. The LUT profile will be a tad
more accurate as you can see yourself when you run the profile test
after the calibration. Especially for near black colors the profile
accuracy is better for the LUT profile, but the matrix profile is
still good there.
Roger wrote:
> Now, it is true that LUT-based profiles are not the end all be all.
> Depends. On good CRTs, a matrix/TRC profile is plenty because there
> is no crosstalk between the channels and the beam amplifier behaves
> as a true radiometric scaler. Butnot so on even the best LCDs.
> Hence, the need for 3D profiles, IMO.
This was true and still is true to some extent. Older flat panel
displays could not be profiled very good with a matrix profile, but
newer and better displays and especially hardware calibrated TFT's
like the Eizo CG21 or the Quato Intelli Proof 21 can be described with
a matrix profile with good success.
Best regards from germany
Peter Karp
Quatographic technology -- www.quato.de
[1] Mac OS loads correction curves to achieve a 1.8 gamma when you
attach a "naked" display without a special calibration. Most
calibration packages for display will use the vcg-Tag to load
calibration curves to the graphics card. There are few which don't
need or use the vcgt. In the iColor software for the Quato Intelli
Proof display the vcgt is used to ensure that _no_ calibration curve
is loaded on the video graphics card, because all calibration is done
in the display itself (whitepoint, gray balance, gamma).
[2] With the Mac OS Colorpicker or a windows tool like Eyedropper or
ColorCop. The monitor was calibrated and profiled to a Gamma of 2.2.
_______________________________________________
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