| Author |
Message |
Roger S.
Guest
|
 Posted:
Wed Oct 19, 2005 2:11 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
"Whatever common knowledge is out there on digital imaging much of it
was created by people with graphics production experience, not digital
imaging hardware/software development experience."
How do you know that, and what's your source of information on "how the
technology really works"?
Scanner hardware, monitors, and contone printers are RGB devices, so
why not work in RGB? Not that I have anything against LAB- use it if
you want to. |
|
| Back to top |
|
 |
Lorenzo J. Lucchini
Guest
|
| Posted:
Wed Oct 19, 2005 2:13 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
mp wrote:
| Quote: | All I can say is that you can *attach* a profile to an image instead of
*applying* it, and this has advantages (wrt quantization errors); I
assume you could do the very same if you (and your profiling >program)
take LAB as the standard color space to work in.
LAB process is much simpler:
1. establish white point/black point
2. render image
|
Excuse me, how are you going to correct whatever color bias you scanner
(or printer, on the other side of the chain) has by doing *just that*?
A profile *must* somehow be involved, whether you work in LAB or not.
| Quote: | Therefore, insisting that everything be done in one color space.
I'm not insisting, I'm making a suggestion. It's a good tool in the
toolbox for sure.
LAB can simplify color. Again, I guess you will wait for Adobe to get
on board before it's deemed "The next big thing."
|
LAB might be quite useful, I'm not trying to imply otherwise, I just
don't know much about it.
| Quote: | Otherwise, surely, doing *anything* 16-bit would be a *complete* waste of time.
High-bit work is most relevent when working monotone images. (more
shades of a single color) Other than archiving the original scan or
perhaps multisampling a shadow-filled scan on a scanner, there's
little technical merit for working with high-bit images.
|
How is multi-sampling on a scanner different from single-sampling (at
16-bit) on another scanner that has less noisy sensors?
Or are you simply saying that *no* scanner has sensors good enough to
get you anything after the 8th bit, and that the *only* way to obtain
something in the 9th bit and beyond is by multi-sampling?
| Quote: | Do whatever you want, but please take this away from the thread:
|
Take away what?
| Quote: | Whatever common knowledge is out there on digital imaging much of it
was created by people with graphics production experience, not digital
imaging hardware/software development experience. If they worked with
anyone on the developer side, they worked with the Marketing dept. Not
engineering. Therefore, much of the common knowledge practice has no
basis in how the technology *really* works.
|
If you say so.
At this point, I'd really like some comments on from someone else on this.
by LjL
ljlbox@tiscali.it |
|
| Back to top |
|
|
mp
Guest
|
| Posted:
Wed Oct 19, 2005 2:59 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
| Quote: | Excuse me, how are you going to correct whatever color bias...
Color management's task is to keep track of what the "real" (device |
dependent) color of an image is supposed to be. With LAB, you always
have the "real" color encoded in the file. (device independent)
FYI: Color bias is normally corrected at the firmware/driver.
| Quote: | Multi-sampling:
Some scanners have the ability to multi-sample. they read the same |
line X number of times, compare results and have some limited
intelligence as to which values to choose.
Scanning at a high-bit, there's not as much sampling going on. So, the
precision goes down a great deal.
I come by all this knowledge working for an OEM a little while ago.
I've been laughed at so many times by the Engineers before they kicked
me out of their offices that I've learned the hard way. |
|
| Back to top |
|
|
Lorenzo J. Lucchini
Guest
|
| Posted:
Wed Oct 19, 2005 3:59 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
mp wrote:
| Quote: | Excuse me, how are you going to correct whatever color bias...
Color management's task is to keep track of what the "real" (device
dependent) color of an image is supposed to be. With LAB, you always
have the "real" color encoded in the file. (device independent)
|
I don't see why this should be a good thing, while I can see at least
one reason (quantization errors) why it may become a bad thing.
| Quote: | FYI: Color bias is normally corrected at the firmware/driver.
|
Not if you don't correct it and just attach the scanner's profile to the
scanned image -- something that can have advantages and disadvantages,
as I see it.
The main advantage is, again, the avoidance of quantization errors when
you'll later have to apply a printer (or whatever) profile.
One disadvantage, as I was discussing with Don, is IMHO that, if you
scan at 8-bit, *applying* (and not just attaching) the scanner's profile
*at scan time* (i.e. internally, which is often done at more than 8-bit)
may (or may not, as Don notes) give better quantized results.
| Quote: | Multi-sampling:
Some scanners have the ability to multi-sample. they read the same
line X number of times, compare results and have some limited
intelligence as to which values to choose.
|
I know what multi-sampling is. As far as I know, though, the "limited
intelligence" is usually just "take the average" (and I don't even think
one is able to do much better than that).
| Quote: | Scanning at a high-bit, there's not as much sampling going on. So, the
precision goes down a great deal.
|
Scanning at a high-bit depth *without multi-sampling*, I suppose you mean.
Yeah, of course, single-sampling will always have less precision than
multi-sampling (unless you don't have enough bits to hold the better,
multi-sampled data).
But what I asked you was: if multi-sampling on scanner A is capable of
producing data that go meaningfully beyond the 8th bit, why shouldn't
single-sampling on scanner B (which, assume, has a better, less noisy
CCD than A) be capable of doing the same?
by LjL
ljlbox@tiscali.it |
|
| Back to top |
|
|
Gordon Moat
Guest
|
| Posted:
Wed Oct 19, 2005 6:45 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
mp wrote:
| Quote: | All I can say is that you can *attach* a profile to an image instead of
*applying* it, and this has advantages (wrt quantization errors); I
assume you could do the very same if you (and your profiling >program)
take LAB as the standard color space to work in.
LAB process is much simpler:
1. establish white point/black point
2. render image
|
Agreed that LaB is much simpler, though at some point an output needs to be created
from an image. Picking an editing space, like Ektaspace, ProPhotoRGB, et al, can
greatly simplify the workflow, though if you have lots of time stick to LaB. There
are also some advantages for certain edits to be done in LaB only, since they
involve essentially imaginary colours, in other words colours that cannot be
recreated in a print.
I make an assumption of printing here because monitors are very limited. I doubt
many would have the latest EIZO on their desk, though another reality is that type
of monitor is just a proofing solution to get near SWOP. A real issue is going from
LaB, or some RGB working space, to some flavour of SWOP, or some sort of CMYK,
HiFiColour, HexaChrome, or custom Pantone combination for printing. With so many
printing possibilities now available, the device independent LaB choice is a good
choice for storage when future printing is not known.
| Quote: |
Therefore, insisting that everything be done in one color space.
I'm not insisting, I'm making a suggestion. It's a good tool in the
toolbox for sure.
LAB can simplify color. Again, I guess you will wait for Adobe to get
on board before it's deemed "The next big thing."
|
I remember several years ago the BruceRGB was all the rage. This was an attempt to
simplify workflow for printing outputs (commercial printing, not desktop inkjet).
Then it seemed that everyone jumped on AdobeRGB (1998), and lately more people
playing around with sRGB. There is a bit of that "flavour of the month" feeling to
all this . . . sort of makes me laugh sometimes.
| Quote: |
Otherwise, surely, doing *anything* 16-bit would be a *complete* waste of time.
High-bit work is most relevent when working monotone images. (more
shades of a single color) Other than archiving the original scan or
perhaps multisampling a shadow-filled scan on a scanner, there's
little technical merit for working with high-bit images.
|
It makes for slightly less destructive further editing of images. Unfortunately
most printing systems and layout software only handle 8-bit images. If someone
knows the output, and the scan needs no changes, then an 8-bit only workflow can
greatly streamline your work. Saving time is not a bad reason for doing 8-bit.
| Quote: |
Do whatever you want, but please take this away from the thread:
Whatever common knowledge is out there on digital imaging much of it
was created by people with graphics production experience, not digital
imaging hardware/software development experience.
|
Sure, check out <http://www.gracol.org> for some of the latest. The idea is that
lots of scanning uses are for things that will be offset printed. Obviously there
is also scientific imaging, and another group of standards and experts. It really
depends upon the end uses for those scans.
| Quote: | If they worked with
anyone on the developer side, they worked with the Marketing dept. Not
engineering. Therefore, much of the common knowledge practice has no
basis in how the technology *really* works.
|
GraCOL is open for submissions of concepts that will improve printed results. If an
engineer or scientist really wanted to make a difference and improve the potential,
then contacting GraCOL is the way to help. The GraCOL standard is new and emerging,
though the intention is to do better than SWOP <http://www.swop.org>.
There are several colour scientists involved in the industry. Whether they are
listened to on everything sometimes comes down to cost and time. With high end
devices, the better concepts and ideas are more implemented than they are in more
general applications.
Time constraints limit commercial results more than any other factor. Those who do
this as a hobby, or without time constraints, have the luxury of achieving better
results.
The popularity of certain writers within the industry (like Bruce Fraser, Seth
Resnik, et al) determines more of how things get done in graphics production than
any science or engineering. If some engineers and scientists were better writers,
then there might be more people doing scanning in a better manner.
Ciao!
Gordon Moat
A G Studio
<http://www.allgstudio.com> |
|
| Back to top |
|
|
Don
Guest
|
| Posted:
Wed Oct 19, 2005 5:34 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
On 18 Oct 2005 14:03:54 -0700, "mp" <mpapet@gmail.com> wrote:
| Quote: | All I can say is that you can *attach* a profile to an image instead of
*applying* it, and this has advantages (wrt quantization errors); I
assume you could do the very same if you (and your profiling >program)
take LAB as the standard color space to work in.
|
You're mixing up your quotes. That's no me!
| Quote: | Therefore, insisting that everything be done in one color space.
I'm not insisting, I'm making a suggestion. It's a good tool in the
toolbox for sure.
|
As I said "horses for courses" and LAB certainly has its advantages,
just like other color spaces have their advantages.
| Quote: | LAB can simplify color. Again, I guess you will wait for Adobe to get
on board before it's deemed "The next big thing."
|
I think this is actually aimed at Lorenzo but...
LAB is the "current" thing. I remember reading over in the Photoshop
group a while back that LAB is the color space Photoshop uses
internally. In other words, regardless of the current color space the
actual editing is applied to the internal LAB representation.
| Quote: | Otherwise, surely, doing *anything* 16-bit would be a *complete* waste of time.
|
That's Lorenzo again! You should only answer one message at a time.
Don. |
|
| Back to top |
|
|
blumesan
Guest
|
| Posted:
Sat Oct 29, 2005 1:12 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
I am always amazed ('though perhaps I shouldn't be) that the longer
these threads get the further they stray from the original subject.
The OP wanted to know how to assign (or embed) a profile to his image
file before importing it into his editing software. I am assuming the
following: that the image file originated from a scanner (this is a
scanner group), and that the profile he wanted to attach was the
scanner output profile. Once this were done the image file could be
imported into the editing software and converted to any working space
of choice.
I am dealing with a similar if not identical problem. I can profile my
scanner using an appropriate target. I can get the scanner to output a
raw untagged file. Fortunately in Photoshop I can import the untagged
file, assign the custom scanner profile and then convert to the working
color space of my choosing. It would however be more efficient if the
scanner profile were embedded in the image file. If you are trying to
implement color management it's not a good idea to have untagged files
floating around. So if one could get the scanner software to embed the
profile when writing the image data to file, this would solve my
problem and that of the OP.
Any suggestions folks??
Mike. |
|
| Back to top |
|
|
Don
Guest
|
| Posted:
Sat Oct 29, 2005 4:46 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
On 29 Oct 2005 01:12:02 -0700, "blumesan" <blumesan@verizon.net>
wrote:
| Quote: | I am always amazed ('though perhaps I shouldn't be) that the longer
these threads get the further they stray from the original subject.
|
Actually, that's the rule... ;o)
Sometimes diversions can be useful when one learns something new
unexpectedly and at other times frustrating when the subject keeps
getting obscured.
| Quote: | The OP wanted to know how to assign (or embed) a profile to his image
file before importing it into his editing software. I am assuming the
following: that the image file originated from a scanner (this is a
scanner group), and that the profile he wanted to attach was the
scanner output profile. Once this were done the image file could be
imported into the editing software and converted to any working space
of choice.
|
That is correct. However, the diversion occurred not about the color
space profile (scanner output) but about the scanner profile itself.
That is to say, the profile which is not attached but *applied* to the
raw data with the intention of correcting for any inherent scanner
bias.
While the former (color space) profile may be useful it can always be
attached afterwards, however, the latter (scanner) profile is of very
limited (if any?) use because the image will have to be edited.
Indeed, a scanner profile can often cause damage which has to be
"undone" in editing afterwards.
| Quote: | I am dealing with a similar if not identical problem. I can profile my
scanner using an appropriate target. I can get the scanner to output a
raw untagged file.
|
The important thing to note is that there are two totally different
profiles here: the scanner profile and the color space profile. They
have nothing to do with each other.
| Quote: | Fortunately in Photoshop I can import the untagged
file, assign the custom scanner profile and then convert to the working
color space of my choosing.
|
Exactly, which is why attaching a color space profile when scanning is
not that important. One can always do it afterwards.
| Quote: | It would however be more efficient if the
scanner profile were embedded in the image file. If you are trying to
implement color management it's not a good idea to have untagged files
floating around. So if one could get the scanner software to embed the
profile when writing the image data to file, this would solve my
problem and that of the OP.
|
It's just a matter of convenience but the key thing is, even if that
is not done at scan time it can always be done afterwards.
Don. |
|
| Back to top |
|
|
blumesan
Guest
|
| Posted:
Sun Oct 30, 2005 8:25 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
Don,
Thanks for your response. I quite agree with what you have said,
except for one statement which seems to fly in the face of color
management principles:
| Quote: | While the former (color space) profile may be useful it can always be
attached afterwards, however, the latter (scanner) profile is of very
limited (if any?) use because the image will have to be edited.
Indeed, a scanner profile can often cause damage which has to be
"undone" in editing afterwards.
|
According to my rudimentary understanding of color management the use
of the scanner profile in conjunction with the working space profile
(and the monitor profile) allows the image to be displayed with colors
as close to those of the original image as possible. Thereby making
editing a bit easier.
Nevertheless I can understand that it may be possible to ignore the
scanner profile when importing a raw image into the editing software.
Simply assign a working RGB profile, carry out whatever color
correction and editing are required and save the file tagged with the
working profile. If in fact the scanner profile can somehow "cause
damage" (something that I have not see suggested before) this may be
the preferred approach. Can you elaborate on this point?
| Quote: | It's just a matter of convenience but the key thing is, even if that
is not done at scan time it can always be done afterwards.
|
In my case, quite correct; I would still like to know how to implement
it. For the OP who is unable to utilize his scanner profile, I gather
you would tell him: "forget it"
Thanks again,
Mike. |
|
| Back to top |
|
|
Don
Guest
|
| Posted:
Sun Oct 30, 2005 6:19 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
On 29 Oct 2005 20:59:19 -0700, "blumesan" <blumesan@verizon.net>
wrote:
| Quote: | Don,
Thanks for your response. I quite agree with what you have said,
except for one statement which seems to fly in the face of color
management principles:
While the former (color space) profile may be useful it can always be
attached afterwards, however, the latter (scanner) profile is of very
limited (if any?) use because the image will have to be edited.
Indeed, a scanner profile can often cause damage which has to be
"undone" in editing afterwards.
According to my rudimentary understanding of color management the use
of the scanner profile in conjunction with the working space profile
(and the monitor profile) allows the image to be displayed with colors
as close to those of the original image as possible. Thereby making
editing a bit easier.
Nevertheless I can understand that it may be possible to ignore the
scanner profile when importing a raw image into the editing software.
Simply assign a working RGB profile, carry out whatever color
correction and editing are required and save the file tagged with the
working profile. If in fact the scanner profile can somehow "cause
damage" (something that I have not see suggested before) this may be
the preferred approach. Can you elaborate on this point?
|
Hi Mike,
Yes, that's the theory. The problem occurs when what's on the film is
not really what one wants in the end product. And that covers pretty
much every scan. In other words, no matter how faithfully the film was
reproduced, the post-processing in an external editor afterwards is
going to change all that.
So it all depends on how much editing one does afterwards (and that
includes editing in scanner software!). If the scan is used "as is"
(all settings at neutral or off) then, yes, using a scanner and film
profiles will produce what's on the film.
However, if one applies a scanner profile but the profile goes in the
opposite direction of any subsequent edits, it's going to make matter
worse, not better!
Best explained with an example. Let's say your scanner has a tendency
to add blue. Applying a scanner profile will, therefore, remove blue.
For the sake of the argument, let's say you have a daylight film shot
without flash at night and indoors. The shot has a distinct yellow
cast. Applying a scanner profile will remove blue and make the scan
appear even more yellow! That may be an extreme case and I used it
only as an illustration, but it should explain what I mean. (Of
course, there are also profiles to compensate for exactly this case
but that means, again, going against the nominal scanner profile.)
Now, please don't get me wrong, in general, a scanner profile does not
really do that much harm (as the above extreme example may suggest)
and in many (most?) cases a scanner profile will actually help. But my
main point is that it's not really essential like monitor and printer
profiles because the effects of a scanner profile will be changed
during post processing (usually quite radically).
| Quote: | It's just a matter of convenience but the key thing is, even if that
is not done at scan time it can always be done afterwards.
In my case, quite correct; I would still like to know how to implement
it.
|
That should be the job of your scanner software. I'm not quite clear
what scanner you have but, for example, using Nikon Scan the option is
in Preferences/Color Management under the RGB tab. In there one can
select what color space to tag to the file.
Don. |
|
| Back to top |
|
|
blumesan
Guest
|
| Posted:
Sun Oct 30, 2005 11:08 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
Don,
Many thanks for your reply. The suggested workflow is quite clear to
me now.
I have a Nikon Supercoolscan 4000ED. Using the Nikonscan software (ver
4.02) I assume that to get a raw (un-messed-with) scan I would turn off
all color management as well as all of the postprocessing options (with
the possible exception of ICE which operates intra rather than post
processing). In this case the software will leave the image untagged.
I can then assign a working space RGB profile when I import the file
into Photoshop.
If I want the software to output a tagged file I have to turn CMS on.
In that case, assuming I understand the manual correctly, the process
is as follows: The software will use one of three built-in profiles
depending on the film type (Color positive, Color negative, Kodachrome)
to translate the raw data into the RGB working space choosen in
Preferences/Color Management, tag the file with the working space
profile and write to disk.
Please let me know if my assumptions are correct.
As my venerable professor used to say "Assumptions are the mother of a
f**k-up"
Cheers,
Mike |
|
| Back to top |
|
|
Don
Guest
|
| Posted:
Mon Oct 31, 2005 9:37 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
On 30 Oct 2005 09:08:53 -0800, "blumesan" <blumesan@verizon.net>
wrote:
| Quote: | I have a Nikon Supercoolscan 4000ED. Using the Nikonscan software (ver
4.02) I assume that to get a raw (un-messed-with) scan I would turn off
all color management as well as all of the postprocessing options (with
the possible exception of ICE which operates intra rather than post
processing). In this case the software will leave the image untagged.
I can then assign a working space RGB profile when I import the file
into Photoshop.
If I want the software to output a tagged file I have to turn CMS on.
In that case, assuming I understand the manual correctly, the process
is as follows: The software will use one of three built-in profiles
depending on the film type (Color positive, Color negative, Kodachrome)
to translate the raw data into the RGB working space choosen in
Preferences/Color Management, tag the file with the working space
profile and write to disk.
|
Hi Mike,
Yes, by turning color management system (CMS) off, you'll be able to
produce an untagged file. However, Nikon Scan actually has a specific
option to produce an untagged file even when CMS is on! Simply select
the "Scanner RGB" color space! The difference between that and turning
CMS off is subtle. When you turn CMS off completely, some Nikon Scan
features are disabled, specifically the LCH editor and Unsharp Mask
(and you have to restart NS!). By using "Scanner RGB" color space
those two remain enabled, but the file is still untagged. (For more
see page 89 in the manual or just search for "Scanner RGB".) However,
if you don't plan to edit in Nikon Scan anyway, all that's academic.
There's also another thing and that's scanning "raw" which in context
of scanning has a specific meaning - very close to my heart! ;o)
Skip if you know this already but scanning "raw" means getting the
most from the scanner (maximum bit depth, etc) and not using any of
the editing features of scanner software. As you state above, ICE -
being at least in part hardware based - should be on when applicable.
Some people also scan in linear gamma (1.0) because that's software
too. I do not, for two reasons. One, gamma 2.2 will be the first thing
I would apply before editing anyway and, two, Photoshop's 16-bit mode
is actually 15-bit (!) while Nikon Scan uses all 16-bits when applying
gamma so it's more accurate.
Anyway, there are several reasons why one would want to scan raw, the
main two being:
1. Scanner software editing is really "second best" and with a limited
subset of editing tools. Also, you're forced to work through the
"preview keyhole" instead of blowing up the image to 1600% like in
Photoshop. Although, as mentioned above, Photoshop's 16-bit mode is
actually 15-bit! :-(
2. Archiving. By scanning raw you get the most the hardware can
deliver and you can thereby "freeze" any future deterioration of film.
You would then burn this to DVD/CD or backup in some other way and
work on a copy. This has many advantages. For example, if at a later
date when you get more proficient in Photoshop or, get a new larger
monitor (i.e. change image size) or, simply don't like the edit, etc.
you can always go back to the "virgin" version without having to
rescan which is not only considerably faster but also easier on the
scanner.
One last hint specific to NikonScan and that's Auto Exposure. It tends
to be quite drastic (clipping) especially for negatives so some people
prefer to turn it off and do the exposure manually. The problem is AE
is very "sticky" and hard to turn off. Not only do you have to turn
off Auto Exposure in *all* places i.e. preview (!!!), single and batch
(both positive and negative!) but you have to exit Nikon Scan *and*
turn the scanner off! To be on the safe side, leave it off for a few
seconds. Only then will Auto Exposure really and truly be off! Do note
that, after restarting NikonScan, if at any time afterwards you click
on Auto Exposure - even accidentally or only for preview - it's on
again and you have the repeat the whole procedure to turn it off.
| Quote: | Please let me know if my assumptions are correct.
As my venerable professor used to say "Assumptions are the mother of a
f**k-up"
|
That's a big assumption he's making! ;o)
Don. |
|
| Back to top |
|
|
blumesan
Guest
|
| Posted:
Tue Nov 01, 2005 3:13 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
Don,
Again thanks for your reply. Very useful information clearly
presented.
You have now convinced me that I should, in future, scan "raw", and
that I can stop worrying about scanner profiles. I have never used the
editing features of the scanner software, much preferring to do it all
in Photoshop.
| Quote: |
One last hint specific to NikonScan and that's Auto Exposure. It tends |
to be quite drastic (clipping) especially for negatives so some people
prefer to turn it off and do the exposure manually
I assume this refers to the exposure of the film during scanning. How
does one go about performing the exposure manually?
And one last question: What about the selection of film type in the
scanner software? I don't see any way to bypass this step and this
must surely interpose some software manipulation of the data between
the "raw" scanned data and what is written to the file. Are you of the
school that recommends scanning color negatives as positives and later
reversing in Photoshop?
Thanks again,
Mike. |
|
| Back to top |
|
|
Don
Guest
|
| Posted:
Tue Nov 01, 2005 8:32 pm Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
On 31 Oct 2005 13:13:58 -0800, "blumesan" <blumesan@verizon.net>
wrote:
| Quote: | One last hint specific to NikonScan and that's Auto Exposure. It tends
to be quite drastic (clipping) especially for negatives so some people
prefer to turn it off and do the exposure manually
I assume this refers to the exposure of the film during scanning. How
does one go about performing the exposure manually?
|
Hi Mike,
Yes, it's determining exposure using Analog Gain and looking at the
histogram. It's no different to what Auto Exposure does only you have
more control. The idea is to keep increasing exposure until the
histogram data starts to touch the right edge. That's the ideal
exposure where nothing is lost.
There is a catch, however, and that's that those highlights (the
brightest areas of the image) which touch the right edge of the
histogram often don't really contain much information (e.g.
reflections). That's why it's common to set some clipping. The usual
values are between 0.3 and 0.5. The reason you do that is to extend
the part of the histogram with useful information.
For example, you may get a histogram with a "mountain" covering the
left half and then a thin line from the middle to the right edge. It's
often safe to clip that thin line and extend the "mountain" to cover
the whole available area thereby maximizing dynamic range of useful
data. But if that thin line contains data you want to keep then
obviously you don't want to clip. It all really depends on image
content.
In theory, Auto Exposure should not clip if you set clipping to 0 but,
it does especially in case of negatives because of other reasons (the
image is inverted and processed, etc).
Finally, another hint. You can check the highlights in Photoshop by
using the histogram, of course. But there's a catch. The combined
display i.e. Luminance is "weighed". It's not just all three channels
added up but it's ~60% green, ~30% red and ~10% blue. This means that
even though Luminance may not clip, once you look at individual
channels they may (and usually do) clip!
| Quote: | And one last question: What about the selection of film type in the
scanner software? I don't see any way to bypass this step and this
must surely interpose some software manipulation of the data between
the "raw" scanned data and what is written to the file. Are you of the
school that recommends scanning color negatives as positives and later
reversing in Photoshop?
|
I haven't made up my mind on that yet. I've been wrestling with my
Kodachromes over the last couple of years because Nikons add that ugly
blue cast. That has literally driven me nuts! I ended up scanning each
slide twice, I call it a "twin scan", once for highlight and once for
shadows. I then combine the two using a program I have written. It's a
long story...
Anyway, since I'm now finally getting to grips with that, pretty soon
I'll have to tackle the negatives. My instinct is to use the Nikon
reversal but set the exposure manually. Other than that, turn
everything else off, as usual.
The thinking is the same as for gamma i.e. reversal would be the first
thing I would have to do in Photoshop after scanning anyway, so why
not let Nikon do it 16-bit instead of Photoshop's 15-bit mode. By
setting exposure manually I'd get around the clipping problem. Since
negatives are compressed anyway (unlike slides) there should be plenty
of dynamic range to be very conservative with exposure.
But I'll really have to run some tests first and then see where that
leads me.
Don. |
|
| Back to top |
|
|
blumesan
Guest
|
| Posted:
Thu Nov 03, 2005 8:25 am Post subject:
Re: color profile embedding (not conversion) utility? |
|
|
Don,
Many thanks for your reply. I really appreciate your taking the time
to answer my posts. I hope you will forgive my obsession with details,
but I really want to get this technique right.
Don wrote:
| Quote: | Yes, it's determining exposure using Analog Gain and looking at the
histogram. It's no different to what Auto Exposure does only you have
more control. The idea is to keep increasing exposure until the
histogram data starts to touch the right edge. That's the ideal
exposure where nothing is lost.
If I understand correctly the procedure (using NikonScan) would be: |
1. Turn off color management, editing function and autoexposure.and
make a Preview scan of the image.
2. Open the Curves and Analog Gain Palettes.
3. For each of the color channels in turn; look at the histogram and
make a ballpark adjustment to the analog gain. Since the effect of the
adjustment is not reflected in the histogram in real time, click on
Redraw.
4. Readjust the analog gain settings as needed, redraw again and
repeat until you are happy with the histogram.
5. Ensure that color management, editing functions and autoexposure
remain off. Gamma may be set at 2.2; autofocus and ICE may be engaged.
Perform final scan.
Thanks again for your help.
Mike. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in