Don, need your help with Nikon Scanner

[blumesan]

Don, if this does manage to catch your attention, please forgive my
approach of putting your name in the subject line. Had I been able to
get your e-mail address, I would have contacted you directly.

I placed the following message in this forum a few days ago, but so far
no constructive response.

<<<"Using the Nikon Coolscan 4000ED and Nikon Scan 4.0.2 software in
standalone mode: If one chooses to use the Color Management System,
one must select a color space profile for the calibrated RGB model from

among 11 profiles presented on the menu. My question is simply this:
Is there some way I can induce the software to use a working space RGB
profile that is not included in the menu? (e.g. ProPhoto or
EktaSpace). Naturally assuming I have the required profile stored
somewhere on my hard disk.

To be perfectly clear, I am well aware that I can open an image in
PhotoShop and assign or convert it to any working space profile of my
choice. I specifically wish to avoid this approach. I am looking for a

method which will cause the Nikon software to translate the raw data
using its (built in) device dependent scannner profile directly into a
device independent working space of my choosing." >>>

Having read many of your previous posts re: scanning with Nikon, and
knowing you have delved as deep as anyone into the workings of these
scanners, I thought you might be able to help.

After having given the problem some thought, I have come up with a
home-brewed workaround which I have not yet tried. Your comments
solicited:

The .icm profiles used by my scanning software are located in the
folder: C:\Program Files\Common Files\Nikon\Profiles. I propose to
select one of the profiles that does appear in the NikonScan menu (e.g.
NKBruce.icm) and move it to another folder. Then I would copy the
ProPhoto.icm file to the Nikon folder and rename it NKBruce.icm.
Thereafter, if I select BruceRGB from the Preferences>Color Management
menu, the software should end up using the ProPhoto profile.

On importing the image file into PhotoShop, I am guessing that the
image will be tagged as being in the BruceRGB space and, assuming I am
using ProPhoto as my default working space, I will get a profile
mismatch warning. At that point I would assign (not convert) the
ProPhoto profile to the image.

Sounds almost too easy. What do you (or any other readers) think?

Mike.

 

[Roy G]

<<<"Using the Nikon Coolscan 4000ED and Nikon Scan 4.0.2 software in
standalone mode: If one chooses to use the Color Management System,
one must select a color space profile for the calibrated RGB model from

among 11 profiles presented on the menu. My question is simply this: profile stored
somewhere on my hard disk.



The .icm profiles used by my scanning software are located in the
folder: C:\Program Files\Common Files\Nikon\Profiles. I propose to
select one of the profiles that does appear in the NikonScan menu (e.g.
NKBruce.icm) and move it to another folder. Then I would copy the
ProPhoto.icm file to the Nikon folder and rename it NKBruce.icm.
Thereafter, if I select BruceRGB from the Preferences>Color Management

Sounds almost too easy. What do you (or any other readers) think?

Mike.

Hi.

I most certainly am not an expert on Scanner SW or Nikon.

I would suspect your work-around might cause you more problems than you
imagine.

Profiles have 2 names. The external File name, and an internal one which can
sometimes be read on one of the Profile Properties Tabs.

Ps seems to use the internal name. Your Scanner might do likewise, so
re-naming the external one, could be a waste of time.

Have you simply tried copying your extra Profile into that Nikon Folder, and
then check to see if it is added to the list. That would be my first move,
but I suppose you have already done so.

Best of luck.

Roy G

 

[tomm42]

You might be asking for trouble, to start off I would just copy the
ProPhoto profile to the ..../Nikon/Profiles folder and see if it shows
up, it may, but I wouldn't rename it. Photoshop also has to have the
ProPhoto listed as a color space, and that has to be the color space of
the pic is opened in. Once a picture's color space is changed it is
either clipped and you can't get those colors back or if the scanning
color space is smaller than viewing colorspace Photoshop doesn't ad any
colors. ie If you have a file scanned in Adobe RGB and the PS color
space is sRGB your file colorspace will be clipped, and changing back
to Adobe RGB won't get your colors back. If a file is from an sRGB
colorspace and it is opened in PS with Adobe RGB selected there won't
be any colors added.

Hope this helps.

Tom

 

[blumesan]

You might be asking for trouble, to start off I would just copy the
ProPhoto profile to the ..../Nikon/Profiles folder and see if it shows
up, it may, but I wouldn't rename it. Photoshop also has to have the
ProPhoto listed as a color space, and that has to be the color space of
the pic is opened in. Once a picture's color space is changed it is
either clipped and you can't get those colors back or if the scanning
color space is smaller than viewing colorspace Photoshop doesn't ad any
colors. ie If you have a file scanned in Adobe RGB and the PS color
space is sRGB your file colorspace will be clipped, and changing back
to Adobe RGB won't get your colors back. If a file is from an sRGB
colorspace and it is opened in PS with Adobe RGB selected there won't
be any colors added.

Hope this helps.

Tom

Naturally the first thing I tried was copying the ProPhoto profile to
the Nikon\Profiles folder. No joy! It did not appear in the NikonScan
preferences menu.

I agree with your remarks about changing (converting) the color space
of a scanned image. In an effort to avoid this approach I am trying to
get the scanner output directly in a color space of my choosing (and
one not provided by the scanner software choices).

It is my opinion that one cannot obtain a truly raw scan (direct
conversion of analog CCD sensor data to digital) from this scanner. At
minimum the data has been converted by the firmware to the scanner's
colorspace using one of the built-in scanner profiles(depending on
which RGB color model one chooses; e.g. Color Positive, Color Negative,
Kodachrome). Now if one had access to that profile (which I do not) or
if I could create a custom profile for my scanner (which I cannot) then
it would be a simple matter to assign that profile to the image and
then convert to any working space profile of my choosing. Simply
assigning an arbitrary working space profile to the image will not
change the data but will change the way it is displayed and
consequently make it difficult or impossible to properly edit the
image. I agree that converting the image to another profile is not
what one should do, and it is precisely for this reason that I am
looking for a way to avoid doing a conversion on the image as it comes
from the scanner.

Here is the result of my test of the workaround mentioned previously:
I removed the profile NKNTSC.icm from the Nikon\Profile folder (since
that is a color space I never use). In the same folder I changed the
name of the EktaSpace (Holmes) profile to NKNTSC.icm. In NikonScan
preferences I selected NTSC (1953) as the color space profile for the
calibrated RGB model. In PS my default color space is Adobe RGB. On
opening the scanned image in PS the profile mismatch dialog appeared
informing me that the color space of the image was "Ekta Space PS 5, J
Holmes.icm"

So far, so good. I will require further experimentation to determine
if the image is truly in the Ekta Space color space.

Cheers/Mike

 

[Don]

Don, if this does manage to catch your attention, please forgive my
approach of putting your name in the subject line. Had I been able to
get your e-mail address, I would have contacted you directly.

No problem! I'm happy to help if I can. Posting here also helps others
reading along as well as getting a second opinion in case I make
mistakes or someone disagrees.

Anyway, I'm a bit swamped today, so I'm afraid I'll tackle the message
tomorrow.

Don.

 

[Don]

I placed the following message in this forum a few days ago, but so far
no constructive response.

<<<"Using the Nikon Coolscan 4000ED and Nikon Scan 4.0.2 software in
standalone mode: If one chooses to use the Color Management System,
one must select a color space profile for the calibrated RGB model from
among 11 profiles presented on the menu. My question is simply this:
Is there some way I can induce the software to use a working space RGB
profile that is not included in the menu? (e.g. ProPhoto or
EktaSpace). Naturally assuming I have the required profile stored
somewhere on my hard disk.

OK, sorry about the delay.

I have to preface this by saying that I don't use Nikon's CMS (Color
Management System) anymore or indeed tag any profiles to my scan.
Actually, I don't even use NikonScan anymore since I wrote my own
scanner program. However, I did do some snooping around (more below) a
while back but that was really "just for fun" and fairly superficial.

Having said that, as you've discovered yourself, just putting the
profile in the relevant NikonScan folder doesn't work. That's because
even though NikonScan profiles are ICC profiles (in theory) Nikon "did
something to them". For more info see below.

Basically, the easiest way is to turn NCM off and then assign the
profile manually afterwards. But I'm assuming you tried that already
and - for whatever reason - that was not an acceptable solution.

To be perfectly clear, I am well aware that I can open an image in
PhotoShop and assign or convert it to any working space profile of my
choice. I specifically wish to avoid this approach. I am looking for a
method which will cause the Nikon software to translate the raw data
using its (built in) device dependent scannner profile directly into a
device independent working space of my choosing." >>>

OK, so you did try that.

Is this just for convenience, or do you have other reasons for wanting
NikonScan to do this?

For example, older versions of Photoshop (including PS6 I use)
actually only process 15-bits when working with a 16-bit image.
NikonScan on the other hand, uses all 16.

Having read many of your previous posts re: scanning with Nikon, and
knowing you have delved as deep as anyone into the workings of these
scanners, I thought you might be able to help.

After having given the problem some thought, I have come up with a
home-brewed workaround which I have not yet tried. Your comments
solicited:

The .icm profiles used by my scanning software are located in the
folder: C:\Program Files\Common Files\Nikon\Profiles. I propose to
select one of the profiles that does appear in the NikonScan menu (e.g.
NKBruce.icm) and move it to another folder. Then I would copy the
ProPhoto.icm file to the Nikon folder and rename it NKBruce.icm.
Thereafter, if I select BruceRGB from the Preferences>Color Management
menu, the software should end up using the ProPhoto profile.

That substitution is clever but unfortunately there's a problem. I
tried something similar once, but the other way around. Namely, using
NikonScan profiles in Photoshop. It didn't work.

The problem is that NikonScan profiles are "proprietary". Even though
they are nominally ICC, Photoshop doesn't recognize them. So I took
them apart and implemented a "fix". Do note that I was only concerned
with the "NK*" profiles because I was trying to figure out how the
Kodachrome mode worked.

The reason PS refuses to recognize "NKLS50.icm" and friends, is
because they have an "unknown" device class designation. For LS50
Nikon switched to "nkpf" which is, apparently, not recognized by the
ICC. Once I (hex) edited this into "scnr" PS was happy to take them.

So, if you have a hex editor you may want to try and change this
device class designation in your ProPhoto profile to "nkpf" and see if
this will fool NikonScan into accepting it.

I have to add that I'm not certain what will happen internally! This
may pretend to "work" (i.e. be accepted by NikonScan") but not really
work i.e. it will not have the desired effect.

The reason I say that is because after I performed the above
modification of the "NKLS50.icm" and friends I was able to apply them
in Photoshop but the effect there were very "mild". Nothing like what
comes out of NS4. So I suspect these profiles have a certain "bias"
which NikonScan compensates for and Photoshop doesn't.

On importing the image file into PhotoShop, I am guessing that the
image will be tagged as being in the BruceRGB space and, assuming I am
using ProPhoto as my default working space, I will get a profile
mismatch warning. At that point I would assign (not convert) the
ProPhoto profile to the image.

Sounds almost too easy. What do you (or any other readers) think?

As I say I did not explore this thoroughly but my conclusion was that
in addition to the profiles the Twain driver itself "did things" to
the scan. So the profiles are really just the icing on the cake.

For what is worth my educated guess (but still a guess) is that, as
far as scanner profiles for my LS50 go it works like this. Each time a
scan is performed (and I found this out by taking apart the Preview
file) *two* profiles are attached (assuming NCM is on).

One:
NKLS50.icm
is always present and I suspect that the scanner profile.

The other depends on various settings. My guess (from names) is that:

NKLS50_K.icm - Kodachrome
NKLS50_MN.icm - monochrome
NKLS50_N.icm - negative
NKLS50_P.icm - positive
NKLS50_R.icm - anyone's guess? Reverse?

Do note this is all from memory because I've archived the results of
my investigation to CD and I don't have my archives handy right now.


The bottom line is that I suspect you could "convince" NikonScan to
accept the ProPhoto profile and indeed even do it correctly but I
don't think it's worth the trouble. The time and effort (reverse
engineering) is probably just too time consuming.

If you do decide to try this and are technically minded a couple of
interesting links. The ICC specs (including file description) can be
had here:

http://www.color.org/icc_specs2.html

The same site also contains a very handy program:

http://www.color.org/membersonly/profileinspector.html

to inspect the profiles and it will even let you export all the
profile data as text (for analysis, inclusion in other programs, etc.)

Good luck and let us know if you discover anything interesting!

Don.

 

[blumesan]

Don,

Many thanks for your detailed reply. It seems you may have missed my
subsequent post in this thread (#4), or you had no comments to offer.
Anyhow I persist in my efforts.

You wrote:

[["Basically, the easiest way is to turn NCM off and then assign the
profile manually afterwards. But I'm assuming you tried that already
and - for whatever reason - that was not an acceptable
solution...............Is this just for convenience, or do you have
other reasons for wanting NikonScan to do this?"]]

I have rejected this approach for two reasons. The first reason is one
you alluded to in your response:

[["Each time a scan is performed (and I found this out by taking apart
the Preview file) *two* profiles are attached (assuming NCM is on).:

NKLS50.icm
is always present and I suspect that the scanner profile. The other
depends on various settings. My guess (from names) is that:

NKLS50_K.icm - Kodachrome
NKLS50_MN.icm - monochrome
NKLS50_N.icm - negative
NKLS50_P.icm - positive"]]

I agree; by "attached" I assume you mean these profiles are used to
convert the raw data to whatever color space is used when the image is
written to file. But I believe these profiles are employed, even when
NCM is *off*. I reach this conclusion based on the fact that, even
with NCM off, if color negative is selected in the film type menu, when
a color negative is scanned the resulting image appears as a positive.
Further, with NCM off, the image does not have the appearance that one
would expect by viewing raw data, even gamma corrected.

The second reason is, even if I scan with NCM off and, following your
suggestion, assign a profile manually afterwards; what profile would I
choose? It certainly wouldn't be correct to assign e.g. EktaSpace. I
have no clue in which space the image currently resides, but it damn
sure isn't any space for which I have a profile. If I did have that
mystery profile then the solution would be to simply assign that
profile and then convert to the working space profile of my choice.

OK, an alternate approach would be to choose one of the RGB spaces
offered in NCM (e.g. Adobe RGB) and, after opening the image in PS,
convert to my preferred working color space. It seems to me that
converting to a different color space, particularly from a smaller
gamut to a larger, is a sure way to corrupt the data.

Thus I conclude that the only way to obtain the scanned image in a
color space other than the choices offered by NCM is to "trick" the
scanner into using an alternate destination space. You will note from
my second post that the trick apparently worked, but confirmation is
pending. I think this worked because the destination RGB color spaces
offered by NCM are all matrix type profiles (not LUT) as are the
profiles for ProPhoto and EktaSpace. In any case, not having the time,
patience or talent to write my own scanner software, this is the best I
can do.

I would certainly not attempt to mess with the NKLS50 profile and
friends (in my case NKLS4000LS40.icm) since these are 16 bit LUT and I
wouldn't know where to apply them if I could make the needed
modification.

Thanks again for your reply. I very much appreciate your contributions
to this forum and I believe other readers may obtain some benefit from
this discussion. Further comments most welcome.

Mike.

 

[Don]

[["Each time a scan is performed (and I found this out by taking apart
the Preview file) *two* profiles are attached (assuming NCM is on).:

NKLS50.icm
is always present and I suspect that the scanner profile. The other
depends on various settings. My guess (from names) is that:

NKLS50_K.icm - Kodachrome
NKLS50_MN.icm - monochrome
NKLS50_N.icm - negative
NKLS50_P.icm - positive"]]

I agree; by "attached" I assume you mean these profiles are used to
convert the raw data to whatever color space is used when the image is
written to file. But I believe these profiles are employed, even when
NCM is *off*. I reach this conclusion based on the fact that, even
with NCM off, if color negative is selected in the film type menu, when
a color negative is scanned the resulting image appears as a positive.
Further, with NCM off, the image does not have the appearance that one
would expect by viewing raw data, even gamma corrected.

Yes, that's correct. These are "low level" profiles (my term). By
"attached" I mean they are listed in the profile scan itself (one of
the "chunks" in this file). From that I assume that the data in the
(profile) scan is actually raw and that these two profiles "should be
applied" at a later date i.e. the profile file is only tagged with
them.

However, it may very well be (since this is all proprietary Nikon
stuff) that by listing these two profiles Nikon indicates which
profiles *have* been applied! This doesn't make much sense, but one
never knows...

Now then, since I only did this "for fun" and didn't really explore it
fully I don't know if the same happens to the actual TIF once the full
scan is performed. I'm assuming (but don't know this for a fact) that
by the time the full scan is performed these profiles are actually
applied i.e. the data is changed. (Only then is the color space
profile *tagged* as far as I can tell.)

Now, this confirms what you said earlier that getting truly raw data
directly from the CCDs in NikonScan is not easy. At the very least it
appears that "NKLS50.icm" is applied no matter what. Another indirect
confirmation of this is that even with everything turned off, the data
examined with a 16-bit histogram reveals some "irregularities".
Specifically, (on my particular LS-50 at least) the red channel levels
off at about 20-30 "clicks" *below* the maximum and never goes above
that value. (The actual value escapes me now, but I can check later if
you want.) The other two channels do go all the way to the top.

In other words, even if I boost exposure causing massive clipping, for
the red channel this clipping occurs slightly below the maximum value.
That's weird and I'm assuming may have something to do with the
scanner profile but again this is only a guess.

Finally, and this is the joker in the pack, as I mentioned last time
Nikon TWAIN also "does things" and I have a strong suspicion (an
educated guess) that a lot happens in firmware within the scanner
itself before the data even leaves the scanner!

One indirect hint of this is how single-pass multi-scanning is
handled. On my LS-50 this is disabled and I wondered if I could turn
it on. So I hacked the TWAIN and, sure enough, the item suddenly
appeared in the menu! However, setting multi-scanning to anything
other than 1x resulted in an error. So it appears much more goes on in
the firmware than meets the eye. I have the developer kit from Nikon
and the documentation confirms that a lot goes on in the firmware
although they kept quiet about how multi-scanning is handled.

But all that's "too much information"... The point is it does appear
that getting really, truly raw data from the CCDs in NikonScan is not
easy.

The second reason is, even if I scan with NCM off and, following your
suggestion, assign a profile manually afterwards; what profile would I
choose? It certainly wouldn't be correct to assign e.g. EktaSpace. I
have no clue in which space the image currently resides, but it damn
sure isn't any space for which I have a profile. If I did have that
mystery profile then the solution would be to simply assign that
profile and then convert to the working space profile of my choice.

It is my understanding that if only the "low level" scanner profiles
are applied what you end up with is basically raw data. That is to say
"raw" in as much as this scanner-profile corrected data can be called
raw. But at least (and in theory) those low level profiles (which are
apparently applied and not tagged!) have compensated for any scanner
induced errors.

However, as you say, the question then is how do you convert this raw
data into a destination profile. In other words, you need a source
profile for reference (i.e. what am I converting *from*. I think,
given all the uncertainties, that the lesser of all evils may be to
try and assign the widest gamut profile available in NikonScan. In
other words, let NikonScan handle this conversion from whatever the
source is i.e. the "from" gamut, to this widest possible gamut. Then
later convert from that (known) gamut to your ProPhoto gamut.

Of course, the first thing that comes to mind is that this will
inevitably cause some data corruption no matter how minor. Namely,
instead of a single conversion (from the "mystery" source gamut to
your desired destination gamut) it will be a two step process. Mystery
gamut to wide gamut, and then wide gamut to ProPhoto.

OK, an alternate approach would be to choose one of the RGB spaces
offered in NCM (e.g. Adobe RGB) and, after opening the image in PS,
convert to my preferred working color space. It seems to me that
converting to a different color space, particularly from a smaller
gamut to a larger, is a sure way to corrupt the data.

Not exactly. If (in this context) you're referring to the "mystery"
gamut as smaller and, say, Adobe RGB as larger then there will be no
data corruption aside from small rounding errors. But the important
thing is the number of colors will not be reduced.

Namely, what will happen is that the data in the smaller gamut would
be "spread out" into the wider gamut. So no data (i.e. colors) will be
lost as would be case if you were doing it the other way around.

However, some small rounding errors may happen simply because no two
gamuts are evenly spaced in integer steps. For example, let's assume
(for illustration only) that the smaller gamut original data is in,
can only handle 100 colors. Now if the destination gamut can handle,
say, 130 colors, you will not be losing any color data i.e. the 100
colors can be represented on 130-color scale but (after conversion)
the distance between individual colors will not be equally spaced.

Of course, if you were doing it the other way around, then there would
be some loss of information because one can't squeeze 130 colors into
100 "slots" without occasionally having to jam two different colors
into 1 slot.

Thus I conclude that the only way to obtain the scanned image in a
color space other than the choices offered by NCM is to "trick" the
scanner into using an alternate destination space. You will note from
my second post that the trick apparently worked, but confirmation is
pending. I think this worked because the destination RGB color spaces
offered by NCM are all matrix type profiles (not LUT) as are the
profiles for ProPhoto and EktaSpace. In any case, not having the time,
patience or talent to write my own scanner software, this is the best I
can do.

Yes, that may work too, but I would do some meticulous testing to
confirm that there is no loss of data. Or, more precisely, that any
potential loss of data is not significant i.e. within the parameters
of your requirements. In my experience this is best done with a good
histogram (which is why I wrote a true 16-bit histogram instead of the
Photoshop's 8-bit approximation). Speaking of which:

http://www.reindeergraphics.com/free.shtml

page contains a free "Wide Histogram" Photoshop plug-in. It doesn't go
all the way (only to 10 bits?) but that's still better than the usual
8-bit approximation most software uses.

I would certainly not attempt to mess with the NKLS50 profile and
friends (in my case NKLS4000LS40.icm) since these are 16 bit LUT and I
wouldn't know where to apply them if I could make the needed
modification.

The ICC profile inspector I mentioned last time would let you export
this as data. But as I also mentioned I don't think that's enough
because of all the other complications i.e. both NikonScan and the
firmware seem to "do things" on top of the profile.

Thanks again for your reply. I very much appreciate your contributions
to this forum and I believe other readers may obtain some benefit from
this discussion. Further comments most welcome.

You're most welcome! Not a lot of people care for such level of detail
but now it's all out there and, who knows, someone else may jump in.

Mind you, even though I'm very picky about all this I do at times pull
back and ask myself if I'm going too far in my quest for perfection
and, indeed, often cut some corners! But I prefer to do that knowing
that I'm cutting corners and choosing which corners to cut, instead of
just blindly trusting the software to make that choice for me.

I guess that makes me a "control freak" but for me that's a badge of
honor I wear with pride i.e. a compliment as in: I know what I'm
doing! ;o)

Don.

 

[blumesan]

Don,

Once again thanks for your most interesting and detailed reply.

It is my understanding that if only the "low level" scanner profiles
are applied what you end up with is basically raw data. That is to say
"raw" in as much as this scanner-profile corrected data can be called
raw. But at least (and in theory) those low level profiles (which are
apparently applied and not tagged!) have compensated for any scanner
induced errors.

However, as you say, the question then is how do you convert this raw
data into a destination profile. In other words, you need a source
profile for reference (i.e. what am I converting *from*. I think,
given all the uncertainties, that the lesser of all evils may be to
try and assign the widest gamut profile available in NikonScan. In
other words, let NikonScan handle this conversion from whatever the
source is i.e. the "from" gamut, to this widest possible gamut. Then
later convert from that (known) gamut to your ProPhoto gamut.

This make a good deal of sense. Since the scanner "knows" what the
source (mystery) profile is, let the scanner make the conversion into a
known profile. This is exactly the logic behind my attempt to trick
the scanner into using EktaSpace (or ProPhoto) as the destination
profile. If this succeeds then I avoid any need for a conversion, even
from a wide gamut to a somewhat smaller one. As I mentioned, my file
name change stunt has apparently worked. At least the image saved to
file is tagged with the name of desired profile. Whether or not it
really is in that color space remains to be determined.

Yes, that may work too, but I would do some meticulous testing to
confirm that there is no loss of data. Or, more precisely, that any
potential loss of data is not significant i.e. within the parameters
of your requirements. In my experience this is best done with a good
histogram (which is why I wrote a true 16-bit histogram instead of the
Photoshop's 8-bit approximation).

To test this, I plan the following. Make two scans of the same slide,
one selecting the EktaSpace profile (disguised by renaming) and a
second using a standard NCM provided profile like sRGB. I have had the
reindeergraphics wide histogram plugin installed for some time. Using
this tool I will look for differences in the histograms of the two
images. Can you suggest any particular features of the histogram that
should be scrutinized?

Mind you, even though I'm very picky about all this I do at times pull
back and ask myself if I'm going too far in my quest for perfection
and, indeed, often cut some corners! But I prefer to do that knowing
that I'm cutting corners and choosing which corners to cut, instead of
just blindly trusting the software to make that choice for me.

I quite agree with that sentiment. I am not so adverse to trusting the
software to make certain decisions, as long as I know what those
decisions are. I don't mind flying on autopilot as long as the GPS
system is functioning to tell me where I am. Then I can always find my
way home.

Thanks again,

Mike.
..

 

[Don]

Don,

Once again thanks for your most interesting and detailed reply.

Hi Mike,

No problem. At a risk of sounding schmaltzy but I've gained so much
from this group so it's only fair to try and give some of it back.

As I mentioned, my file
name change stunt has apparently worked. At least the image saved to
file is tagged with the name of desired profile. Whether or not it
really is in that color space remains to be determined.

One interesting test may be the following:

Scan an image and save as NEF. That's supposed to as raw as NikonScan
can do it.

Next, open this NEF file in NikonScan and save it *twice* as TIF:
1. Using your "fudged" profile
2. Using any standard NikonScan profile.

Finally, strip away the TIF headers and only export raw data (you can
do this is Photoshop in the "Save As" dialog). Make sure you use the
same setting for both files (e.g. byte order, etc).

Now, in theory, if the profile is truly only tagged those two files
(containing just image data) should be identical.

If they are not, then NikonScan "does things" at some very low level.

If they are identical, then you could tag your profile afterwards.

Actually to be absolutely sure you may want to save a 3rd file from
the NEF "using scanner profile" i.e. without tagging any file.

Again, the image data contained there should be identical to the other
two if NikonScan doesn't mess with it at some low level.

To test this, I plan the following. Make two scans of the same slide,
one selecting the EktaSpace profile (disguised by renaming) and a
second using a standard NCM provided profile like sRGB. I have had the
reindeergraphics wide histogram plugin installed for some time. Using
this tool I will look for differences in the histograms of the two
images. Can you suggest any particular features of the histogram that
should be scrutinized?

I would just look at the actual numerical data. You will notice some
"oddities" simply because of the way Wide Histogram reduces 16-bit
information. This often causes recursive and overlapping "waves" also
known as "gappy" histograms. (That's why I implemented another way of
reducing this data to get a smooth histogram.)

Anyway, you can definitely see the differences more clearly in the
Wide Histogram, if for no other reason than because you can actually
look at the numerical data and even import it into Excel, etc.
Photoshop histogram can show this data too but only one at a time if
you hover over each histogram "bin" which is far too cumbersome.

Also, the trouble with multiple scanning is that the image will be
different no matter how small those differences are. Therefore, the
above procedure by using NEF as an intermediate stage assures you
always start with the same data which makes it much easier to compare.

Don.

 

[blumesan]

Don,

Just a few questions to clarify your suggestions:

One interesting test may be the following:

Scan an image and save as NEF. That's supposed to as raw as NikonScan
can do it.

I assume you mean that this scan must be performed with NCM *off*.

Next, open this NEF file in NikonScan and save it *twice* as TIF:
1. Using your "fudged" profile
2. Using any standard NikonScan profile.

Finally, strip away the TIF headers and only export raw data (you can
do this is Photoshop in the "Save As" dialog). Make sure you use the
same setting for both files (e.g. byte order, etc).

I assume this means: open the TIF files in PS and then perform a "Save
As". Do you mean simply uncheck the profile box and thus save as an
untagged image (no embedded profile)? This would strip away the
profile tag, but even without an embedded profile the image data would
remain unchanged and in their original color space (whatever that may
have been).

Now, in theory, if the profile is truly only tagged those two files
(containing just image data) should be identical.

How identical? They should now reside in two different color spaces.
It seem you are suggesting that when the NEF file is saved as a TIF
using the two profiles, the profiles are simply *assigned* to the image
file and that no color space conversion (into the selected destination
space) takes place. If this is the case then we are back to the image
residing in the "mystery" profile. The whole object of the excersize
was to get the Nikon software to convert the image from the mystery
profile into a known destination space.

If they are not, then NikonScan "does things" at some very low level.

If they are not, I would suspect that the "raw" NEF file data has
indeed been converted into the color spaces specified in the save
operation. And this is what we want (no?).

If they are identical, then you could tag your profile afterwards

Presumably they are already tagged, even if not converted.

Actually to be absolutely sure you may want to save a 3rd file from
the NEF "using scanner profile" i.e. without tagging any file.
Again, the image data contained there should be identical to the other
two if NikonScan doesn't mess with it at some low level.

Same questions as above.

I would just look at the actual numerical data. You will notice some
"oddities" simply because of the way Wide Histogram reduces 16-bit
information. This often causes recursive and overlapping "waves" also
known as "gappy" histograms. (That's why I implemented another way of
reducing this data to get a smooth histogram.)

Anyway, you can definitely see the differences more clearly in the
Wide Histogram, if for no other reason than because you can actually
look at the numerical data and even import it into Excel, etc.
Photoshop histogram can show this data too but only one at a time if
you hover over each histogram "bin" which is far too cumbersome.

Also, the trouble with multiple scanning is that the image will be
different no matter how small those differences are. Therefore, the
above procedure by using NEF as an intermediate stage assures you
always start with the same data which makes it much easier to compare.

That part is clear. It make sense to start with the same image in
order to make valid comparisons. But if all 3 NEF files after saving
as TIF are identical (as you suggest), what does that tell me about my
approach to getting the NCM to convert the "raw or mystery" data into a
color space of my choosing?

Sorry to trouble you with the additional questions. Hope you don't
tire of the subject.

Mike.

 

[blumesan]

Don,

A quick addendum. Following the steps you suggested:

Scan an image and save as NEF. That's supposed to as raw as NikonScan
can do it.

Next, open this NEF file in NikonScan and save it *twice* as TIF:
1. Using your "fudged" profile
2. Using any standard NikonScan profile.

After opening the NEF files in NikonScan and attempting to save as TIF,
there is no option available (that I can see) to use or attach any
profile when saving. That's as far as I got with preliminary testing.

Mike

 

[blumesan]

Ok Don, have a look at these: Since I was not able to follow your
procedure for reasons stated in last post, this required 4 separate
scans:

#1 NCS on. Selected profile in NikonScan: Ekta Space (using renamed
file as described above). Profile identified in PS as Ekta Space.

#2 NCS on. Selected profile in NikonScan: sRBG. Profile identified
in PS as sRGB.

#3 NCS off. Saved directly to file as TIF. Profile identified in PS
as sRGB. I have read elsewhere that this is an idiosyncrasy of
NikonScan, to embed the sRGB profile when NCM is turned off.

#4 NCS off. Saved directly to file as NEF. Reopened in NikonScan
and saved as TIF. Profile identified in PS as sRGB.

There certainly are differences between #1, #2 and #3, though I'm
damned if I know what to make of them. #3 and #4 are identical, as one
would expect. Somewhat surprising to me is that the EktaSpace file
(which is a much wider gamut than sRGB) has the narrowest range in the
histogram.

Curiouser and curiouser.

Mike

http://i10.photobucket.com/albums/a143/blumesan/histogramtest2.jpg

 

[Don]

I assume you mean that this scan must be performed with NCM *off*.

Yes, to be on the safe side.

In theory though (i.e. just to compare the two files) it shouldn't
make any difference. In order to test if the two files are the same
the NCM status of the source doesn't really matter as long as it's the
same in both cases, of course.

But just to be on the safe side it's best to turn NCM off.

I assume this means: open the TIF files in PS and then perform a "Save
As". Do you mean simply uncheck the profile box and thus save as an
untagged image (no embedded profile)? This would strip away the
profile tag, but even without an embedded profile the image data would
remain unchanged and in their original color space (whatever that may
have been).

No, I mean, strip away the whole TIF header (PS actually adds a footer
as well!).

You do use "Save As..." for this but then go to Format, select file
type "Raw (*.RAW)" and click on [Save]. This will bring up another
dialog with "Raw options". Leave "Header" at "0". The "Interleaved
order" and "Byte order" don't really matter (for our purpose) as long
they are the same for both (or all three) test files. You will get an
alert box with a warning that "some data will not be saved" (that
refers to the header, etc) but since that's exactly what we want just
click on [OK].

Of course, if you want to import this raw data then you should make
note of "Interleaved order", image size (width, height), bit depth and
possibly "Byte order" (in case of 16-bit images). These values are
then supplied if you want to import this raw data back into Photoshop.

When I was playing around with this I would then compare these files
in DOS with: "fc /b file1 file2". If the DOS box fills up with data
and starts scrolling the files are different. If nothing happens (it
takes a while) the command will come back with something like "no
differences encountered". In Windows you can use "Windiff" but that's
only installed with some MS programming languages. I'm sure there are
also various fancy other 3rd party compare programs out there.

How identical? They should now reside in two different color spaces.
It seem you are suggesting that when the NEF file is saved as a TIF
using the two profiles, the profiles are simply *assigned* to the image
file and that no color space conversion (into the selected destination
space) takes place.

Yes, that's my suspicion. The file is "tagged" with a generic profile
rather than "converted". As far as I was able to tell only the
"native" profiles I listed last time (scanner, KC, negative, etc.) are
actually applied to the data (i.e. the data is converted). The other
"generic" profiles I suspect are only tagged. In Photoshop, of course,
you have much more freedom because you can specify how a profile is to
be handled. I believe you get 3 choices.

If this is the case then we are back to the image
residing in the "mystery" profile. The whole object of the excersize
was to get the Nikon software to convert the image from the mystery
profile into a known destination space.

That's correct, but by doing this test we can check if this "mystery"
color space makes any difference. If the data is the same, then there
is no mystery color space and you can tag the image freely. You can do
that either in NikonScan (by fudging i.e. replacing an exiting
profile) or simply by tagging a new profile in PS.

If they are not, I would suspect that the "raw" NEF file data has
indeed been converted into the color spaces specified in the save
operation. And this is what we want (no?).

Actually it can probably mean a couple of things depending on what
NikonScan really does. But if they are different I would deduce that -
at the very least - NikonScan has some "awareness" of the "mystery"
(source) color space and applies this knowledge to image data before
tagging it with the desired destination profile.

Presumably they are already tagged, even if not converted.

No, that means the only conversion that takes place is when applying
those "native" Nikon profiles (scanner, KC, negative, etc.). After
that, the "generic" profiles are only tagged but the actual image data
(already modified by "native" profiles) remains the same.

That part is clear. It make sense to start with the same image in
order to make valid comparisons. But if all 3 NEF files after saving
as TIF are identical (as you suggest), what does that tell me about my
approach to getting the NCM to convert the "raw or mystery" data into a
color space of my choosing?

It would imply that either, any such conversion takes place before the
profile is even known or, no such explicit conversion takes place at
all but is incorporated in the scanner profile, for example. Either
way, that in turn means it shouldn't really matter whether the profile
is tagged in NikonScan or afterwards because the actual image data is
the same in either case.

Sorry to trouble you with the additional questions. Hope you don't
tire of the subject.

No, problem, at all Mike. I myself had the same learning curve trying
to figure out what the KC mode does so I know what you're going
through. Each answer just opens up new questions.

As I said back then, I've done quite a few things in my life but
nothing was as complex as optics. It's like those Russian dolls, every
time you open one, there's another one inside...

Don.

 

[Don]

After opening the NEF files in NikonScan and attempting to save as TIF,
there is no option available (that I can see) to use or attach any
profile when saving. That's as far as I got with preliminary testing.

I was writing from memory so I may be wrong, but I thought using a NEF
file was the same as scanning. Perhaps this only applied to editing
tools (like curves, etc)?

Unfortunately I only have NikonScan 3.1 on this machine and no
scanner... :-/

I'll try and run some tests tomorrow on another machine with NS4. The
trouble I won't have access to my scanner in the next few days.

Don.

 

[Don]

Ok Don, have a look at these: Since I was not able to follow your
procedure for reasons stated in last post, this required 4 separate
scans:

#1 NCS on. Selected profile in NikonScan: Ekta Space (using renamed
file as described above). Profile identified in PS as Ekta Space.

#2 NCS on. Selected profile in NikonScan: sRBG. Profile identified
in PS as sRGB.

#3 NCS off. Saved directly to file as TIF. Profile identified in PS
as sRGB. I have read elsewhere that this is an idiosyncrasy of
NikonScan, to embed the sRGB profile when NCM is turned off.

#4 NCS off. Saved directly to file as NEF. Reopened in NikonScan
and saved as TIF. Profile identified in PS as sRGB.

There certainly are differences between #1, #2 and #3, though I'm
damned if I know what to make of them. #3 and #4 are identical, as one
would expect. Somewhat surprising to me is that the EktaSpace file
(which is a much wider gamut than sRGB) has the narrowest range in the
histogram.

Weird as it sounds but I can conceive a couple of scenarios where that
may occur. Another problem is that when you use a wide gamut beyond
the capabilities of the monitor Photoshop adds its own "fudging" by
often approximating colors which it can't display. (That's why I never
trust my eyes when trying to get a handle on things like this and only
rely on data.)

BTW, this Photoshop "fudging" is the easiest to observe when you blow
up a 16-bit image to maximum magnification. If you look carefully
you'll see that virtually all pixels are *not* "pure" (of one color)
but actually "hatched" using two (or more) different colors.

If memory serves, in art (i.e. in painting) that's called
"pointillism" when colors are not mixed physically but "optically".
When observed at close range you can see individual points of
different color, but when you pull back the points merge into a single
color. This "optical mixing" is said to result in much more vibrant
colors than if the composite paints were mixed physically.

But I'm off on another digressionary tangent... ;o)

The point (sic) is, artists have a choice and do this intentionally,
Photoshop on the other hand has not choice but uses this approach as
"second best" to approximate more colors than the hardware is capable
of.

Curiouser and curiouser.

Yup, that's the way it was for me too trying to figure out what KC
mode does!

Don.