Computer Graphics with Photoshop - Gamut Warning Problem

Author

Gamut Warning Problem

andre.gunther@gmail.com

2005-06-16, 7:17 pm

Hello,

i have some pictures that I would like to have printed. I have
donwloaded a color profile from the printshop of my choice. I can see
some relatively large areas where Photoshop shows me a Gamut warning
for that profile.
I can get rid of this by reducing the saturation of the image, but I
was wondering if there is a better way to do this.
I am looking for some transformation that keeps the overall appearance
of the image without having to reduce saturation.
If you know how to handle this, I would highly appreciate your input.

Thanks

Andre
http://www.aguntherphotography.com

Mike Russell

2005-06-16, 7:17 pm

andre.gunther@XXXXXXXXXX wrote:
> Hello,
>
> i have some pictures that I would like to have printed. I have
> donwloaded a color profile from the printshop of my choice. I can see
> some relatively large areas where Photoshop shows me a Gamut warning
> for that profile.
> I can get rid of this by reducing the saturation of the image, but I
> was wondering if there is a better way to do this.
> I am looking for some transformation that keeps the overall appearance
> of the image without having to reduce saturation.
> If you know how to handle this, I would highly appreciate your input.

It's hard to know what you mean by "relatively large". Rather than changing
the entire image, try using the sponge tool to desaturate the particular
areas a bit. If there is a particular area - for example a very saturated
blue object, you may have better luck adjusting the color in curves, and
adding contrast in the black or complimentary (aka unwanted) color to at
least retain texture and tonal variation where color was lost.

In general, use Relative Colorimetric, if you're not already, to preserve
the colors that do fit withink the gamut.

If the profile is public, post the URL and some of us will take a look.
--

Mike Russell
www.curvemeister.com

andre.gunther@gmail.com

2005-06-16, 11:17 pm

Hi Mike,

thanks for your answer. It makes sense. By relatively large I mean
large enough, so that one would prob. see posterization when printed
this way.
I am worried about large areas in my images being clipped.

I am using the ezprints.com profile. Its available on their website.
They are cheap and offer at least a profile.

I have always used perceptual so far for conversions, since I did my
editing in Adobe RGB and I wanted to preserve the appearance.
If I use Relative Colorimetric, don't I change the appearance of the
image and won't I get the same posterization artefacts I mentioned
above, since other colors don't change accordingly?
If I compress the colorspace, shouldn't I do it for the whole image?
Does it mean that EZprints can not print saturated colors?

As for my workflow, I am doing all the editing in AdobeRGB and before
saving i convert to sRGB. The more I think about it, I am not sure this
is a wise thing though. Care to comment?
Ezprints also wants to have sRGB files. So I am using the ezprints.icc
profile for proofing colors only. I am not sure that is correct, but I
have never converted to the ezprints profile. Am I doing the right
thing here?
Also when I use Proof Colors, do I see exactly what the print will look
like?

Well, I still have a lot to learn. I appreciate all the help I can get.

On a somewhat related note. I read on outbackphoto a while ago about
ProRGB color space. Someone was praising it, since it is an even wider
color space than AdobeRGB. At the end, in order to print it, don't you
have to compress the color space anyways? So what is the point in doing
all your editing in a larger color space?

Andre
http://www.aguntherphotography.com

Mike Russell

2005-06-16, 11:17 pm

andre.gunther@XXXXXXXXXX wrote:
> Hi Mike,
>
> thanks for your answer. It makes sense. By relatively large I mean
> large enough, so that one would prob. see posterization when printed
> this way. I am worried about large areas in my images being clipped.

Gamut clipping may not be serious, provided there is luminance variation you
may not notice it at all. Using the sponge tool to desaturate some areas
and then using RelCol intent is often better than remapping all the colors
in the image as is normally the case for perceptual.

> I am using the ezprints.com profile. Its available on their website.
> They are cheap and offer at least a profile.

Yes, I took a look at their profile, and it appears to be of good quality.
The shadows in RelCom have a bit more red added to them than in perceptual,
which is a bit surprising.

> I have always used perceptual so far for conversions, since I did my
> editing in Adobe RGB and I wanted to preserve the appearance.
> If I use Relative Colorimetric, don't I change the appearance of the
> image and won't I get the same posterization artefacts I mentioned
> above, since other colors don't change accordingly?

The answer depends on your particular image. You can get pretty good results
by relying on perceptual to remap your out of gamut colors by mapping the
entire image to fit within the smaller gamut of the print, but this amounts
to "dumbing down" all the colors of your image, and generally results in
duller colors.

Relcol allows most of your colors to be brighter, and takes the risk of some
colors being mapped to the edge of the gamut, preserving luminance. You can
control that risk by desaturating only those colors until they are within
gamut.

> If I compress the colorspace, shouldn't I do it for the whole image?

It depends on the image. If you do this for all your images you'll get an
average result, but if you want some colors that sing, even if it means some
clipped colors, RelCol is generally better.

> Does it mean that EZprints can not print saturated colors?
Their color gamut looks about like any other printer's.

> As for my workflow, I am doing all the editing in AdobeRGB and before
> saving i convert to sRGB. The more I think about it, I am not sure
> this is a wise thing though. Care to comment?

You've touched on an area where reasonable people disagree. My own
preference is to do everything in sRGB, and the main reason for this is to
avoid accidentally putting an image on the web that loses saturation. Adobe
RGB is a perfectly good color space as well, and there are still others who
work with ProPhotoRGB.

> Ezprints also wants to have sRGB files. So I am using the ezprints.icc
> profile for proofing colors only. I am not sure that is correct, but I
> have never converted to the ezprints profile. Am I doing the right
> thing here?

My recommendation would be to calibrate your monitor carefully, either with
Adobe Gamma or a device like the Spyder or ColorOne, and send your prints
as-is in sRGB. Send some test prints that include very saturated colors,
and judge for yourself whether you are happy with the results or not.

> Also when I use Proof Colors, do I see exactly what the print will
> look like?

Not really, but it will be reasonably close. Proofing is based in part on a
standard observer in a standard viewing environment, which is considerably
brighter and bluer than normal light.

> Well, I still have a lot to learn. I appreciate all the help I can get.

Every piece of knowledge you learn will improve your images. My main advice
is to trust what you can see, or measure, and not rely too much on mere
advice or authority, particularly when it involves spending your money :-)

> On a somewhat related note. I read on outbackphoto a while ago about
> ProRGB color space. Someone was praising it, since it is an even wider
> color space than AdobeRGB. At the end, in order to print it, don't you
> have to compress the color space anyways? So what is the point in
> doing all your editing in a larger color space?

I couldn't have said it better myself.

BTW - you've touched on several topics that overlap my article called "The
Mystery of the Blue Glass". I'd be interested in your reaction to it. It
describes an unsolved mystery (though I have some thoughts) involving the
difficulty of printing dark, saturated blues using several color spaces, and
a particular printer profile:
http://www.curvemeister.com/tutoria...orZoo/index.htm

---
Mike Russell
www.curvemeister.com

andre.gunther@gmail.com

2005-06-17, 7:15 pm

Hello Mike,

thanks again for your detailed answer.
I will need to experiment with some of the proposed techniques myself
to find out what works best for me. Somehow I was hoping for a miracle
button or so, that uses a specific printer profile and makes my image
look good on that printer.
The way I understood RelCol Map to work is that only the out of gamut
colors are mapped back to fit the specific color space. If I have a
large range of colors outside (who knows how much really lies outside),
this can lead to posterization, as they are all clipped to the border
of the color space.
Anyways, I guess the right thing to do is play around a bit and figure
out what works best for me.
I have calibrated all my monitors with the Spyer2 (I returned Spyder
after finding out it doesn't work well with LCD). I am mainly working
on my laptop for convenience, even though the display really sucks
(Compaq V2000). I have calibrated it though and it should be reasonably
close. If I keep a constant viewing angle it should be allright.
Interestingly, the out of Gamut colors will stand out when I view my
laptop screen from below (tilting the display back all the way).
This came as a surprise to me and seems to indicate that those colors
really need some work. I have a reddish-yellowish bright patch. The out
of gamut warning shows for the whole patch and when I tilt my screen
the patch is all I can see at some extreme angles.
As for your little puzzle ;-) :
Is the printer a CMYK printer? Have you viewed the result in different
lighting conditions (e.g. fluorescent, daylight).
I don't understand enough about printing, my best guess is this:
A CMYK printer has to mix blue out of some other colors. For a dark
blue, there is prob. a RGB of 0 0 30 or something like that. Now if you
have to mix this out of different colors, each color in the mix will be
even more subdued and spread out (the printer does not make bright or
dark pixels) among the black (to get it dark).
So maybe one out of 50 pixels will have a color value.
Depending on your paper, that may even bleed into the black and vanish
or be noticable as "random" color specks.
Like I said I have no idea what really happens. Seems like printing
technology has yet to catch up to modern camera technology.
I also noticed some posterization in the bright blue areas (or is it
just me?).

On a related note, where can I get ProPhotoRGB profiles? I am not sure
I saw it in Photoshop.

Also I like your website and your tutorials. Seems like I should give
Labmeter a try, although I have a little trouble understanding these
color space representations. I am not sure if yours is different, but
generally they map a 3D color space onto a 2D display and thats where
the confusion starts for me.

Andre
http://www.aguntherphotography.com

Mike Russell

2005-06-17, 7:15 pm

<andre.gunther@XXXXXXXXXX> wrote in message
news:1119028913.487446.107150@g14g2000cwa.googlegroups.com...
> Hello Mike,
>
> thanks again for your detailed answer.
> I will need to experiment with some of the proposed techniques myself
> to find out what works best for me. Somehow I was hoping for a miracle
> button or so, that uses a specific printer profile and makes my image
> look good on that printer.

I like to compare today's printer technology to watches. Right now our
"color watches" are accurate to five or ten minutes a day. Pretty darn
good. If you want to strive for greater accuracy, you can either tweak
Adobe Gamma (like the "fast/slow" adjustment of a watch) every once in a
while, and set your color watch manually for important images, which is my
current approach - or try to get a watch that is more accurate. This costs
money, and may or may not pan out. There is a lot of fluff out there right
now, and much money is being spent on profiles that introduce banding into
your image.

Furthermore, and this is the most important, complete color accuracy reaches
a point of dimiishing returns, and manual adjustment will generally yield a
better result than any automatic process, however well calibrated. I think
the greatest rewards, in terms of actual image appearance, come from knowing
how to do good color correction, and the Curvemeister plugin is dedicated to
this goal. As Hecate said about sharpening, connecting your brain to what
you are doing is the more important thing.

> The way I understood RelCol Map to work is that only the out of gamut
> colors are mapped back to fit the specific color space. If I have a
> large range of colors outside (who knows how much really lies outside),
> this can lead to posterization, as they are all clipped to the border
> of the color space.

Yes, the color is mapped to the closest matchine hue and saturation,
preserving luminance. This is generally a good compromise because the
variation in luminance renders the hue clipping unnoticeable. The trick is
to anticipate when this will not work, and use desaturate or some other
technique to compensate for it.

> Anyways, I guess the right thing to do is play around a bit and figure
> out what works best for me.

Certainly - or at least try to verify the things you read here and in other
forums, and judge for yourself whether it is meaningful. We are at an early
point in the evolution of color correction and color calibration as applied
to photography. This makes it an exciting time, because there is so much
new out there, but it also means we have to weed out the stuff that sounds
good, but doesn't really pan out.

> I have calibrated all my monitors with the Spyer2 (I returned Spyder
> after finding out it doesn't work well with LCD). I am mainly working
> on my laptop for convenience, even though the display really sucks
> (Compaq V2000). I have calibrated it though and it should be reasonably
> close. If I keep a constant viewing angle it should be allright.

If you have a CRT lying around, try plugging it in and see how the colors
compare. In particular, make sure that you are able to distinguish colors
that are different on the CRT, and that you don't have an overall color
cast.

> Interestingly, the out of Gamut colors will stand out when I view my
> laptop screen from below (tilting the display back all the way).
> This came as a surprise to me and seems to indicate that those colors
> really need some work. I have a reddish-yellowish bright patch. The out
> of gamut warning shows for the whole patch and when I tilt my screen
> the patch is all I can see at some extreme angles.

OOG colors will tend to show up because you are driving one of the display
channels at its brightest value. Changing the angle of your LCD tends to
blot out the darker values more abruptly, because LCD's are based on a light
polarization phenomenon.

> As for your little puzzle ;-) :
> Is the printer a CMYK printer? Have you viewed the result in different
> lighting conditions (e.g. fluorescent, daylight).

No - it's an RGB inkjet. I'm thinking of changing the puzzle to invite
people to try printing the image on their own printers. Blue cobalt is a
devil of a color to print - the glass turns opaque so easily.

> I don't understand enough about printing, my best guess is this:
> A CMYK printer has to mix blue out of some other colors. For a dark
> blue, there is prob. a RGB of 0 0 30 or something like that. Now if you
> have to mix this out of different colors, each color in the mix will be
> even more subdued and spread out (the printer does not make bright or
> dark pixels) among the black (to get it dark).
> So maybe one out of 50 pixels will have a color value.
> Depending on your paper, that may even bleed into the black and vanish
> or be noticable as "random" color specks.

What you are saying is accurate. The question remains whether we can
actually analyze what is happening to the soft proof stage first, and then
take that further to the actual printed image. This article is a work in
progress, and I would like to invite others to share their ideas and
thoughts.

> Like I said I have no idea what really happens. Seems like printing
> technology has yet to catch up to modern camera technology.
> I also noticed some posterization in the bright blue areas (or is it
> just me?).

Possibly - it's internal glass reflections, which can look like
posterization.

> On a related note, where can I get ProPhotoRGB profiles? I am not sure
> I saw it in Photoshop.

I believe it is downloadable from
http://www.dodgecolor.com/downloads.html

> Also I like your website and your tutorials. Seems like I should give
> Labmeter a try, although I have a little trouble understanding these
> color space representations. I am not sure if yours is different, but
> generally they map a 3D color space onto a 2D display and thats where
> the confusion starts for me.

Thanks for the kind words. LabMeter shows a slice of the color space. Open
up the curve adjustment layer and move the whtie end of the Lightness curve
up and down, and you'll see different slices at each lightness value.
--
Mike Russell
www.curvemeister.com

andre.gunther@gmail.com

2005-06-17, 7:15 pm

Hi Mike,

thanks again for the answer. Does anyone know how to cite passages when
writing the messages online (possibly through my gmail account?). I
don't want to cut and past as this means I have to put the ">"
manually.

Back to the topic, I wish there was a photoshop plugin/action/button
that would let me increase saturation of my image while intelligently
remaining within a specified color profile. Even though it would result
in some compression (e.g. logarithmic) one could still achieve pretty
good results.
I don't have that much time and I prefer shooting over retouching. I
therefore try to automate as much of my workflow as I can (hence the
need for an action ;-) )
I am using the CRT only when I am getting stuff printed. Often I just
play around or prepare images for the web. I don't take this too
seriously. But when I spend a lot of money for some poster size prints,
I want them to come out correct the first time (without having to order
smaller prints first to check out the appearance).
Anyways, I have learned something new today, thats what counts.
Thanks for the link.

The whole color management problem is one fuzzy nebular thing in my
head. To my understanding a pixel has 256 red, 256 green and 256 blue
values (for 8 bit RGB). So whats the deal, how can a pixel be too red
to be printed?
How can I have 8bit Adobe RGB and 8bit sRGB and yet the Adobe can store
more color information? Each pixel stores exactly 24bits of color (in
both color spaces).

Andre
http://www.aguntherphotography.com

Mike Russell

2005-06-17, 7:15 pm

news:1119045447.337179.298040@g44g2000cwa.googlegroups.com...
> Hi Mike,
>
> thanks again for the answer. Does anyone know how to cite passages when
> writing the messages online (possibly through my gmail account?). I
> don't want to cut and past as this means I have to put the ">"
> manually.

Stuggling with different mailers right now myself. Can't help you there,
other than to say I'm more likely to actually read something I'm replying to
if I have to format it manually :-)

> Back to the topic, I wish there was a photoshop plugin/action/button
> that would let me increase saturation of my image while intelligently
> remaining within a specified color profile. Even though it would result
> in some compression (e.g. logarithmic) one could still achieve pretty
> good results.

Not a bad idea at all. One quick trick to bumping saturation is to assign a
higher gamut profile, such as Adobe RGB, to an sRGB image. Bumps the colors
right up. My favorite is to make a and b steeper in Lab mode. Neither of
these exactly addresses what you're requesting, an "auto saturation"
adjustment.

> I don't have that much time and I prefer shooting over retouching. I
> therefore try to automate as much of my workflow as I can (hence the
> need for an action ;-) )

I think an action that bumps the saturation would do the job in a practical
way. Keep an eye out for images that don't fit the pattern, and modify them
separately, perhaps with an action with different settings.

> I am using the CRT only when I am getting stuff printed. Often I just
> play around or prepare images for the web. I don't take this too
> seriously. But when I spend a lot of money for some poster size prints,
> I want them to come out correct the first time (without having to order
> smaller prints first to check out the appearance).
> Anyways, I have learned something new today, thats what counts.
> Thanks for the link.

I work that way too. Most of my snapshots are printed straight out of the
camera, but for a large image I'll spend considerable time making it look
good.

> The whole color management problem is one fuzzy nebular thing in my
> head. To my understanding a pixel has 256 red, 256 green and 256 blue
> values (for 8 bit RGB). So whats the deal, how can a pixel be too red
> to be printed?

One way to translate that is that it is too red to be printed at a
distinctly different color from other nearby red colors of the same
brightness.

> How can I have 8bit Adobe RGB and 8bit sRGB and yet the Adobe can store
> more color information? Each pixel stores exactly 24bits of color (in
> both color spaces).

They do, and in fact Adobe RGB devotes some color values, such as
RGB(255,0,0) to colors that, until recently, were impossible to see on a
CRT. In practice, its not a big deal whether you use sRGB and Adobe RGB.

Right now I'm noticing that a lot of the free printer profiles out there,
even the ones provided by the manufacturer, are of poor quality.
--

Mike Russell
www.curvemeister.com