Nikon Color Management and Profile Quality

[hpowen]

Seems I recall discussions here about the quality (or lack thereof) of
Nikon's canned color profile for the Coolscan 4000. The problem, IIRC,
was severely posterized shadow detail when using the supplied profile
and the Nikon CMS. The searches I've performed aren't returning the
info I'm looking for, so I'm hoping someone will jog my memory.

I've been running Nikon Scan with the CMS turned off, and I'm trying to
convince a friend to do the same (as well as ignore the canned
profile). Am I giving bad advice? He seems to think so.

 

[rafe bustin]

Seems I recall discussions here about the quality (or lack thereof) of
Nikon's canned color profile for the Coolscan 4000. The problem, IIRC,
was severely posterized shadow detail when using the supplied profile
and the Nikon CMS. The searches I've performed aren't returning the
info I'm looking for, so I'm hoping someone will jog my memory.

I've been running Nikon Scan with the CMS turned off, and I'm trying to
convince a friend to do the same (as well as ignore the canned
profile). Am I giving bad advice? He seems to think so.

As far as I can tell there isn't much to Nikon's
CMS; all it really does is tag your file with a
chosen standard ICC working space. (I choose
AdobeRGB.)

I'm not sure what you mean by the "canned" profile.

I don't believe I have issues with posterized
shadow detail; that's mostly a matter of properly
setting your black point, and not necessarily
letting the auto-exposure do it for you.


rafe b.
http://www.terrapinphoto.com

 

[John]

As far as I can tell there isn't much to Nikon's
CMS; all it really does is tag your file with a
chosen standard ICC working space. (I choose
AdobeRGB.)

I'm not sure what you mean by the "canned" profile.

I don't believe I have issues with posterized
shadow detail; that's mostly a matter of properly
setting your black point, and not necessarily
letting the auto-exposure do it for you.


rafe b.

As far as I can ascertain, NikonScan uses a proprietary CMS which uses its
own generic built-in (or 'canned') profiles. On my Windows 2000 system,
these reside in Program Files/Common Files/Nikon/Profiles. You will see that
some have underscores in the names - if you try to use these profiles in
Photoshop, they will not work. There is one profile which does work in
Photoshop (the one without an underscore in the name), but I am not happy
with the results it gives - they concurr with those of the OP.

I believe (although I don't know for sure) that the profiles are
non-standard and only work with the NikonScan CMS. It doesn't simply tag the
scan files - it does a conversion from the 'canned' scanner profile to the
specified working space (e.g. Adobe RGB (1998)). At first sight, it seems to
do a reasonable job, but closer inspection shows shadow clipping and
posterisation. There may be an issue with auto-exposure, but the clipping
only happens when the profile is used - i.e. if you select 'Scanner RGB',
even with auto-exposure on, there is no clipping. Therefore I conclude that
the profile is responsible.

The bad news is that NikonScan will not recognise any profiles other than
those provided - you can try dropping custom profiles into the Nikon profile
area but they won't show up in NikonScan CMS. In any case, if they are in a
proprietary format, only Nikon could generate one which would work with
their CMS. Good old Nikon!

 

[Jim]

Seems I recall discussions here about the quality (or lack thereof) of
Nikon's canned color profile for the Coolscan 4000. The problem, IIRC,
was severely posterized shadow detail when using the supplied profile
and the Nikon CMS. The searches I've performed aren't returning the
info I'm looking for, so I'm hoping someone will jog my memory.

I've been running Nikon Scan with the CMS turned off, and I'm trying to
convince a friend to do the same (as well as ignore the canned
profile). Am I giving bad advice? He seems to think so.

Turning off the Nikon scanner CMS is appropriate when you have created your
own scanner profile since the software does not permit selection of such
profile. I got better results that way; I let PS convert from the scanner
space to my working space.
Jim

 

[Don]

I believe (although I don't know for sure) that the profiles are
non-standard and only work with the NikonScan CMS. It doesn't simply tag the
scan files - it does a conversion from the 'canned' scanner profile to the
specified working space (e.g. Adobe RGB (1998)). At first sight, it seems to
do a reasonable job, but closer inspection shows shadow clipping and
posterisation. There may be an issue with auto-exposure, but the clipping
only happens when the profile is used - i.e. if you select 'Scanner RGB',
even with auto-exposure on, there is no clipping. Therefore I conclude that
the profile is responsible.

I second all of that. Together with the previous paragraph it's a very
good summary of how NCM works - or, more accurately, doesn't! ;o)

Also, in an exchange with Nikon "Support" (on another subject,
Kodachromes) I was advised by Nikon themselves to turn NCM off -
permanently! This so-called "support" is not very reliable but even
they seems to have noticed the damage NCM does...

Although some have reported good results with NCM, that doesn't change
the fact of what NCM actually does to the data. And there I agree with
John. Unlike conventional color management systems which simply
instruct the application on how the data is to be interpreted by using
a harmless tag (which can be easily removed or modified later), NCM
seems to actually change the data (and that can *not* be undone later
without major loss of quality!).

This is identical to Photoshop's "convert to profile" rather than "tag
with profile" only NCM doesn't give this choice. Instead it goes
willy-nilly and sprays NCM graffiti all over the data...

In Photoshop, this setting is hiding under "Edit / Color settings..."
on the "Color Management Policies" tab where one can set the defaults
on how untagged files are to be handled.

So, to the OP, the way to convince your friend is as follows (assuming
there were no changes since the last time I, for one, tried NCM -
which was > 2 yrs ago):

Ask him to remove the NCM from a file and then show him how this can
be done in Photoshop without actually changing the file. Also, two
scans in succession, one with and one without NCM could be very
revealing. One caveat though, NikonScan is *very* cranky when asked to
turn something off! So, don't just restart NikonScan but - to be on
the safe side - turn off the scanner and wait for about 1 minute for
all the power to drain!

Don.

 

[John]

On Wed, 18 May 2005 11:38:08 +0100, "John"

So, to the OP, the way to convince your friend is as follows (assuming
there were no changes since the last time I, for one, tried NCM -
which was > 2 yrs ago):

Don't worry Don, nothing has changed :-( Last year, full of (misplaced)
hope, I installed NikonScan 4 instead of the V3.1 that came with the
scanner. Absolutely no improvement and, (the reason I 'upgraded' to 4 in the
first place) the bug which causes NikonScan to spew out a linear gamma file
when ICE is used in conjunction with Nikon CMS, was still there. Sigh :-(

 

[UrbanVoyeur]

Seems I recall discussions here about the quality (or lack thereof) of
Nikon's canned color profile for the Coolscan 4000. The problem, IIRC,
was severely posterized shadow detail when using the supplied profile
and the Nikon CMS. The searches I've performed aren't returning the
info I'm looking for, so I'm hoping someone will jog my memory.

I've been running Nikon Scan with the CMS turned off, and I'm trying to
convince a friend to do the same (as well as ignore the canned
profile). Am I giving bad advice? He seems to think so.

It's probably better than leaving it on, but not by much. You can do
yourself and your friend a gigantic favor and purchase a Wolf Faust
calibration slide (if you are shooting slides) or a calibration card to
shoot if you are shooting color neg.

Then use your favorite calibration capable software (Monaco, VueScan,
Silverfast, etc) to generate your own scanner profile. The difference
will be dramatic.

Even a gray card with white balance would be an improvement.

Also, you might try scanning into the Adobe or BruceRBG color space.
Both have a wider color gamut than sRGB

 

[rafe bustin]

As far as I can ascertain, NikonScan uses a proprietary CMS which uses its
own generic built-in (or 'canned') profiles. On my Windows 2000 system,
these reside in Program Files/Common Files/Nikon/Profiles. You will see that
some have underscores in the names - if you try to use these profiles in
Photoshop, they will not work. There is one profile which does work in
Photoshop (the one without an underscore in the name), but I am not happy
with the results it gives - they concurr with those of the OP.

This is mostly nonsense.

The reason you can't see these files from within
Photoshop is that they need to be (for Win2000)
in the directory

\WINNT\system32\spool\drivers\color

Now as to the assertion that (for example) Nikon's
"AdobeRGB" profile is different from (say) Adobe's
profile of the same name, surf over to

http://www.iccview.de


and give that site a look-see. What they have
there is a great VRML tool that lets you compare
any two profiles in three dimensions. They have
a "stock" of all the standard working space profiles,
and you can upload any profile of your choosing
as well. (I've used the tool to inspect printer
and monitor profiles that I've created.)

Anyway.. what you'll see is that there's no
discernable difference between Nikon's "AdobeRGB"
and the system version.

I stand by my assertion that Nikon's CMS is mostly
innocuous, and that Don's general ranting on this
topic is... well, what Don does.


rafe b.
http://www.terrapinphoto.com

 

[Don]

Don't worry Don, nothing has changed :-( Last year, full of (misplaced)
hope, I installed NikonScan 4 instead of the V3.1 that came with the
scanner. Absolutely no improvement and, (the reason I 'upgraded' to 4 in the
first place) the bug which causes NikonScan to spew out a linear gamma file
when ICE is used in conjunction with Nikon CMS, was still there. Sigh :-(

fx: Shakes head... After several run-ins with Nikon's so-called
"support" nothing surprises me from Nikon anymore... :-(

I was forced to install NikonScan (NS) 4 when I got the new scanner
(LS-50) because NS 3 refused to work with it. Then to add insult to
injury, NS4 wouldn't install on my W98 machine, so I had to install
W2K first! Argh...!!

In the end, I wrote my own program, which - oh, wonder - works just
fine under W98... :-/

Don.

 

[Don]

I stand by my assertion that Nikon's CMS is mostly
innocuous, and that Don's general ranting on this
topic is... well, what Don does.

Would you be so kind as to point to any of this alleged "ranting",
specifically?

Are you saying that NCM does *not* irreversibly modify file data
instead of harmlessly tagging it with a profile?

Are you also saying that effects of NCM on LS-30 when scanning
Kodachromes are "innocuous"!?!?

Etc. etc.

To an impartial observer, above *objective facts* couldn't be any
further from "ranting". If anything, the said observer could only
deduce that the shoe is firmly on the other foot.

Don.

 

[Kennedy McEwen]

I was forced to install NikonScan (NS) 4 when I got the new scanner
(LS-50) because NS 3 refused to work with it. Then to add insult to
injury, NS4 wouldn't install on my W98 machine, so I had to install
W2K first! Argh...!!

Works fine on this W98 machine here.

 

[Kennedy McEwen]

Anyway.. what you'll see is that there's no
discernable difference between Nikon's "AdobeRGB"
and the system version.

I stand by my assertion that Nikon's CMS is mostly
innocuous, and that Don's general ranting on this
topic is... well, what Don does.

Indeed. And should anyone be tempted to investigate further and write
some software to decode the Nikon profiles, as I have, you will find
that they are fully compliant with the ICC specifications - which do
actually permit profiles to be application and peripheral specific.

 

[Kennedy McEwen]

Are you saying that NCM does *not* irreversibly modify file data
instead of harmlessly tagging it with a profile?

Don, its a *PROFILE* not a tag. It isn't supposed to *just* tag the
file!

The profile is a part of the system colour management. It translates
the native colour space of the Nikon scanner into the abstract, but
wider gamut, colour space of AdobeRGB or whatever is selected, and
*then* applies the tag to the file to indicate to any CM aware software
that the image in the file exists in that colour space and not any
other.

It may well be that your problem is actually down to the software you
are using, not NCM. You have frequently mentioned that you are still
using PS6, which didn't exactly have the best of integral CM
capabilities.

 

[John]

This is mostly nonsense.

Oh, is it? Have you tried it then? What you obviously haven't done is read
my post properly. I suggest you do so.

The reason you can't see these files from within
Photoshop is that they need to be (for Win2000)
in the directory

\WINNT\system32\spool\drivers\color

I can assure you I have a thorough knowledge of colour management and how it
works. I know exactly where the profiles need to be stored, which is why I
copied them into that folder to try them in Photoshop. The point I was
making was that they are designed *specifically* for Nikon CMS - that is
*why* they are in a non-standard folder. And as I said, you can't use them
in Photoshop. These profiles have names which end in underscore something,
e.g. NKLS4000LS40_N.icm, NKLS4000LS40_K.icm, etc. (presumably standing for
'negative', 'Kodachrome', etc). There is one exception - a profile which
gets installed in the above folder during installation and is designed
(presumeably) for Windows ICM. This is called (e.g.) NKLS4000LS40.icm, and
*does* work in Photoshop: you can scan raw, assign this profile to your
file, then convert to Adobe RGB (1998). The results are very similar to
Nikon CMS - clipping and all. I have tried all this so I know what I am
talking about. Have you?

In case my original post was not clear, I will try again. Correct me if I am
wrong, but in your post, you suggest that NikonScan CMS simply tags (or
assigns) Adobe RGB (1998) (or whatever is selected) to the raw file, yes?
Well, it doesn't. It assigns a 'canned' scanner profile (one of those in
Program Files/Common Files/Nikon/Profiles as appropriate) to the raw data,
and then performs a *conversion* to the user specified working space. It is
this conversion that damages the data.

You don't believe me? OK, try this test. Scan a slide with Nikon CMS of or
set to Scanner RGB. Open in Photoshop and look at the histogram. Make a note
of this and make sure there is no clipping. Of course, you can assign any
profile you like to the file, it won't affect the data. No surprises there.

Now select Adobe RGB (1998) in Nikon CMS as your working colour space. Scan
the slide again. Open in Photoshop and compare the histogram. It will be
different - very different! What does this tell us? That a profile to
profile conversion has taken place, that's what.

Now as to the assertion that (for example) Nikon's
"AdobeRGB" profile is different from (say) Adobe's
profile of the same name

Where did I say that?

and give that site a look-see. What they have
there is a great VRML tool that lets you compare
any two profiles in three dimensions. They have
a "stock" of all the standard working space profiles,
and you can upload any profile of your choosing
as well. (I've used the tool to inspect printer
and monitor profiles that I've created.)

Anyway.. what you'll see is that there's no
discernable difference between Nikon's "AdobeRGB"
and the system version.

Like I said, you've completely misunderstood my post. It is the 'canned'
scanner profile that causes the problem, not Nikon's version of AdobeRGB
(1998). I am fully aware that there is no diference between it and the
system version.

I stand by my assertion that Nikon's CMS is mostly
innocuous, and that Don's general ranting on this
topic is... well, what Don does.

Don is not alone here. Have a look at
http://www.creativepro.com/story/review/14539.html
This is a review of the Coolscan 4000 by Bruce Fraser from a while ago. Read
what he says - perhaps he talks nonsense as well.

 

[John]

This is mostly nonsense.

Oh, is it? Have you tried it then? What you obviously haven't done is read
my post properly. I suggest you do so.

The reason you can't see these files from within
Photoshop is that they need to be (for Win2000)
in the directory

\WINNT\system32\spool\drivers\color

I can assure you I have a thorough knowledge of colour management and how it
works. I know exactly where the profiles need to be stored, which is why I
copied them into that folder to try them in Photoshop. The point I was
making was that they are designed *specifically* for Nikon CMS - that is
*why* they are in a non-standard folder. And as I said, you can't use them
in Photoshop. These profiles have names which end in underscore something,
e.g. NKLS4000LS40_N.icm, NKLS4000LS40_K.icm, etc. (presumably standing for
'negative', 'Kodachrome', etc). There is one exception - a profile which
gets installed in the above folder during installation and is designed
(presumeably) for Windows ICM. This is called (e.g.) NKLS4000LS40.icm, and
*does* work in Photoshop: you can scan raw, assign this profile to your
file, then convert to Adobe RGB (1998). The results are very similar to
Nikon CMS - clipping and all. I have tried all this so I know what I am
talking about. Have you?

In case my original post was not clear, I will try again. Correct me if I am
wrong, but in your post, you suggest that NikonScan CMS simply tags (or
assigns) Adobe RGB (1998) (or whatever is selected) to the raw file, yes?
Well, it doesn't. It assigns a 'canned' scanner profile (one of those in
Program Files/Common Files/Nikon/Profiles as appropriate) to the raw data,
and then performs a *conversion* to the user specified working space. It is
this conversion that damages the data.

You don't believe me? OK, try this test. Scan a slide with Nikon CMS of or
set to Scanner RGB. Open in Photoshop and look at the histogram. Make a note
of this and make sure there is no clipping. Of course, you can assign any
profile you like to the file, it won't affect the data. No surprises there.

Now select Adobe RGB (1998) in Nikon CMS as your working colour space. Scan
the slide again. Open in Photoshop and compare the histogram. It will be
different - very different! What does this tell us? That a profile to
profile conversion has taken place, that's what.

Now as to the assertion that (for example) Nikon's
"AdobeRGB" profile is different from (say) Adobe's
profile of the same name

Where did I say that?

and give that site a look-see. What they have
there is a great VRML tool that lets you compare
any two profiles in three dimensions. They have
a "stock" of all the standard working space profiles,
and you can upload any profile of your choosing
as well. (I've used the tool to inspect printer
and monitor profiles that I've created.)

Anyway.. what you'll see is that there's no
discernable difference between Nikon's "AdobeRGB"
and the system version.

Like I said, you've completely misunderstood my post. It is the 'canned'
scanner profile that causes the problem, not Nikon's version of AdobeRGB
(1998). I am fully aware that there is no diference between it and the
system version.

I stand by my assertion that Nikon's CMS is mostly
innocuous, and that Don's general ranting on this
topic is... well, what Don does.

Don is not alone here. Have a look at
http://www.creativepro.com/story/review/14539.html
This is a review of the Coolscan 4000 by Bruce Fraser from a while ago. Read
what he says - perhaps he talks nonsense as well.

 

[rafe bustin]

Like I said, you've completely misunderstood my post. It is the 'canned'
scanner profile that causes the problem, not Nikon's version of AdobeRGB
(1998). I am fully aware that there is no diference between it and the
system version.

I guess my question is, what reason do you have
to expect that Nikon's "canned" profiles (the ones
in the Nikon folder, not the system folder) would
be of any use or interest to you?

Clearly, any scanner is going to have to do some
sort of transformation of its raw RGB data in
order to fit that data to one of the standard
working spaces. But that transformation is
going to be proprietary by its very nature, and
it's not reasonable to expect it to be open,
transparent, or user-accessible.

In other words... I've scanned thousands of
images on my LS-8000 over these last four years
or so... and really never took the slightest
interest in the "profiles" in the Nikon folder.

I hear lots of bellyaching about NikonScan,
which I find a bit amusing. Not to put too
fine a point on it, but NikonScan is the best
scanner driver I've ever seen or used, and
I've used a few.

Now as to the other point that you and Bruce
Fraser make about compressed or posterized
shadow detail, I'll refer you to this URL
and the scans in the table labeled "One
Perfect Chrome." (Scroll down to near the
bottom of the page.)

http://www.terrapinphoto.com/jmdavis/

IMHO, the Nikon holds its own against
some stiff competition here in resolving
shadow detail, on a very tough bit of image.
You will find no posterization in the
Nikon scan snippet.


rafe b.
http://www.terrapinphoto.com

 

[rafe bustin]

To an impartial observer, above *objective facts* couldn't be any
further from "ranting". If anything, the said observer could only
deduce that the shoe is firmly on the other foot.

Clearly you and John and I are dealing with
a different set of "objective facts."

I don't pretend to know what goes on inside of
NikonScan, nor am I presumptuous enough to write
my own scanner driver or image-processing
applications. What I know is that I'm quite
content with the program, having used it
productively over a period of almost exactly
four years now on my LS-8000.

I respect Bruce Fraser's knowledge of color
management (and own a copy of his book on the
subject) but I don't take his reviews of
specific scanners or scanner drivers as the
last word.

To tell the truth, I didn't bother with ICC
color management until about two years ago,
but when I finally made the move, nothing
much changed in my scanning routines or the
images produced by my scanner.

On the few occasions where I've had to deal
with Nikon service (in the few weeks after
I purchased my scanner) they were helpful
and cooperative. Maybe I'm just lucky.


rafe b.
http://www.terrapinphoto.com

 

[John]

I guess my question is, what reason do you have
to expect that Nikon's "canned" profiles (the ones
in the Nikon folder, not the system folder) would
be of any use or interest to you?

Because if they adhered to the same standard as Adobe et al, it would be
possible to substitute a custom profile if I wasn't happy with the canned
one. I agree that if they worked satisfactorily, there would be no need to
worry about them. The only reason for going into that level of detail was to
refute your comment that I was talking nonsense. I was also trying to
correct your mistaken assertion that NikonCMS simply assigns a working
space.

For me, Nikon CMS has not worked satisfactorily in all cases. Having said
that, I have had excellent results as well - it depends on the image and
what is critical. Don't forget also that you are using a superior scanner to
myself and Don - maybe the deficiencies of Nikon CMS are less significant in
your case.

Clearly, any scanner is going to have to do some
sort of transformation of its raw RGB data in
order to fit that data to one of the standard
working spaces. But that transformation is
going to be proprietary by its very nature, and
it's not reasonable to expect it to be open,
transparent, or user-accessible.

But it *is* reasonable to want to use custom profiles. Nikon's approach does
not allow for this. Yes, there is a way round - turn off Nikon CMS, scan raw
and assign a custom profile in Photoshop. But wasn't this the original
question - 'is it better to turn off Nikon CMS?' and 'are the canned
profiles any good?'

In other words... I've scanned thousands of
images on my LS-8000 over these last four years
or so... and really never took the slightest
interest in the "profiles" in the Nikon folder.

I hear lots of bellyaching about NikonScan,
which I find a bit amusing. Not to put too
fine a point on it, but NikonScan is the best
scanner driver I've ever seen or used, and
I've used a few.

Now as to the other point that you and Bruce
Fraser make about compressed or posterized
shadow detail, I'll refer you to this URL
and the scans in the table labeled "One
Perfect Chrome." (Scroll down to near the
bottom of the page.)

http://www.terrapinphoto.com/jmdavis/

IMHO, the Nikon holds its own against
some stiff competition here in resolving
shadow detail, on a very tough bit of image.
You will find no posterization in the
Nikon scan snippet.

I won't dispute your personal experience, but at least you should concede
that others (like myself) have had problems with it. Dismissing the comments
of those who have spent some time trying to evaluate the exact nature of the
problem (this is after all a 'scanners' newsgroup, not a 'photo' one) as
'mostly nonsense' is unhelpful if not insulting

 

[Philip Homburg]

Because if they adhered to the same standard as Adobe et al, it would be
possible to substitute a custom profile if I wasn't happy with the canned
one.

And there are good reasons for making your own profiles. The built-in
profile can take into account only the characteristics of the scanner and
not of the slide film or the kind of light you are using.

With complex issues like this one, a separation of concerns is good. Let
NikonScan convert a slide to pixels, and use another program to perform
color management.

 

[rafe bustin]

I won't dispute your personal experience, but at least you should concede
that others (like myself) have had problems with it. Dismissing the comments
of those who have spent some time trying to evaluate the exact nature of the
problem (this is after all a 'scanners' newsgroup, not a 'photo' one) as
'mostly nonsense' is unhelpful if not insulting

I still think it's nonsense to expect
that you can make use of the profiles
in the Nikon folder. You asserted that
these are not "true" ICC profiles, and
I have shown that at least some are.

I have never been interested in profiling
my film scanner, and so Nikon's CMS, such
as it is, has been perfectly adequate.

It is a bit odd that the Nikon CMS needs
to be turned off to support custom scanner
profiles, but this really comes down to
a matter of aesthetics.


rafe b.
http://www.terrapinphoto.com