Scanner Profiling: Critical or Luxury?

[Alan Smithee]

How important is it to profile the scanner? If the image is ultimately going
into Photoshop for final editing and color corrections (on a calibrated
monitor) does the scanner profile really matter?

 

[Lorenzo J. Lucchini]

How important is it to profile the scanner? If the image is ultimately going
into Photoshop for final editing and color corrections (on a calibrated
monitor) does the scanner profile really matter?

Quite a subjective question you're asking.

My view: a color profile of the scanner will give you scans that are
"perfect" as color is concerned, and within the limitations of the
scanner and the profiling technique.

So, profiling is certainly important if you want to reproduce faithfully
whatever you've put in the scanner.

If, however, you're going to edit colors later in Photoshop, then you
aren't going to want "perfect" colors: you want the colors you like. In
theory, this would imply that you shouldn't care about profiling.

In practice, personally, I'd still like to have a "perfect-colors" image
saved even if I plan to make edits, simply because my eye isn't skillful
enough to say "ok, these are the colors I wanted", but rather needs
something to compare to -- and the more faithful to the original picture
is that something, the better.


If you've got a calibrated monitor, and only care about getting the
colors *you* like, and are skilled enough to maintain consistent color
judgement, than I think you'll be fine with no scanner profiling.

Also, remember that color correction is a destructive operation, and
with 8-bit images this fact is important in practice. So, depending on
the way your software works, you may *not* want to perform *two* color
corrections (one by the scanner driver, based on profiling, and one by
yourself, based on your judgement).


by LjL
(e-mail address removed)

 

[Jim]

How important is it to profile the scanner? If the image is ultimately
going
into Photoshop for final editing and color corrections (on a calibrated
monitor) does the scanner profile really matter?

You always use a profile with a scanner. That is the means to associate
colors with sensor readings.

Sometimes, you use the one that the manufacturer supplies. Other times, you
use the one that you create.
The profile that you create can be more accurate for your particular scanner
that the one that comes with the unit.

Yes, the scanner profile matters unless you really have too much hair on
your head.
Jim

 

[Don]

How important is it to profile the scanner? If the image is ultimately going
into Photoshop for final editing and color corrections (on a calibrated
monitor) does the scanner profile really matter?

It depends on what you're after.

In theory (!) a scanner profile is supposed to eliminate scanner bias
and produce an image identical to what's on the media.

However, it does that by changing data which comes from the scanner.
Now, most people would probably say "Good! That's what I want!". But,
others would say: "Not so fast, Skippy!"... ;o)

For one, in some cases a scanner profile can actually do a lot of
damage e.g. when the unprofiled biased scan is closer to the desired
output so in editing you have to undo what scanner profile has done.

Also, there are contrarians aka the "raw scan sect" ;o) who believe
that any "corruption" no matter how well meaning is still corruption.

Put another way, you *can't* remove the scanner profile without losing
some data, but you *can* apply the scanner profile to a raw scan and
be largely none the wiser, i.e. a scan with the scanner profile on is
comparable to a raw scan with the scanner profile applied afterwards.

And since the end product is rarely identical to what's on the film
this added flexibility is certainly something to consider. In other
words, the scanned image (whether scanner profiled or not) will be
edited. And the prime directive of image processing to limit the
number of edits.

Don.

 

[Andy Spragg]

Yes, the scanner profile matters unless you really have too much hair on
your head.

A figure of speech evidently - but not one I'm familiar with. Can you
explain? Is it some sort of allusion to being simian in nature?

 

[bmoag]

The manufacturer supplies a profile that is automatically installed with the
driver.
You can buy test transparency images for profiling but if you think about it
all you have profiled is that particular emulsion exposed and processed
under those particular conditions. That profile will not necessarily be any
more accurate than the generic profile provided by the scanner manufacturer
for any other transparency, particularly one made on a different emulsion
processed differently.
Negative materials for profiling are even more complex and not readily
avaiable, although you can make your own as you can for transparencies.
In truth monitor profiling in a color managed environment is accurate enough
for most uses but you have to be careful about your work habits and
recalibrate your monitor regularly.
Custom paper/printer profiling can significantly improve the process if you
are willing to put the money, time and effort into the process but,
depending on what you are after, may not be worth the effort.

 

[Don]

A figure of speech evidently - but not one I'm familiar with. Can you
explain? Is it some sort of allusion to being simian in nature?

The premise is that not using a scanner profile would cause so much
frustration that it would lead to loss of hair.

There are two possible things he's alluding to:

One is referring to an expression often used to express one's
frustration i.e. "I'm pulling my hair out" indicating one has reached
such a stage of frustration as to become deranged (some deranged
people pull their hair out).

The other is that worry and frustration lead to loss of hair i.e.
boldness. Strictly speaking this is not correct because permanent
boldness is inherited (from one's maternal grandfather).

Don.

 

[Andy Spragg]

The premise is that not using a scanner profile would cause so much
frustration that it would lead to loss of hair.

Aha. Now I see. The only premise I could come up with was that for a
scanner profile not to matter, the user would have to be more ape than
man ...

 

[Don]

Aha. Now I see. The only premise I could come up with was that for a
scanner profile not to matter, the user would have to be more ape than
man ...

That's another way of looking at it and maybe that's what he meant, in
which case (as someone who does not use a scanner profile) I would
fall on the simian side, especially with my full head of hair. ;o)

Don.

 

[false_dmitrii]

How important is it to profile the scanner? If the image is ultimately going
into Photoshop for final editing and color corrections (on a calibrated
monitor) does the scanner profile really matter?

Scanners are said to be less variable than monitors and printers. If
the scanner was shipped with good color performance, you might see only
minor inaccuracies in its output for most of its life.

I don't have the practical experience to say whether profiling works
properly in practice. It's used by quite a few people who know what
they're doing, though. In theory...once the image is in a known color
environment such as color-managed Photoshop, it's as good or bad as it
appears. If you can correct it to your satisfaction, the lack of a
profile only "matters" as much as the time you spend on corrections and
your insistence on absolute fidelity to the original colors.

However, if you can scan a reflective or slide-based target under the
exact conditions and (non-adaptive!) settings you'll be using to scan
the image, then use it to create your profile, the profile *should*
allow your final image to come closer to the original colors than you
can accomplish by hand. Contrary to Don's assertion, practically
speaking it's no more "destructive" to apply a profile than to make
your own corrections, particularly if you're working with a 16-bit
file. Rather, you're allowing the software to perform the equivalent
of your own color corrections, only based on accurate measurements
rather than your subjective judgment. *If* the profile does its job,
it's a better way of getting colors on your screen that match the
original media. Whether you want the colors from the original media is
another question.

All of this depends on high-quality profile targets, consistent
hardware performance, and software that does what it's supposed to. If
what comes out on the screen looks nothing like what went into the
scanner, theory falls apart.

false_dmitrii

 

[Don]

Contrary to Don's assertion, practically
speaking it's no more "destructive" to apply a profile than to make
your own corrections, particularly if you're working with a 16-bit
file.

Actually it is! This is why:

When you make your own corrections you can always go back to your
backup/archive and start again.

Once the scanner has *applied* (not assigned!) the profile there's no
way to get the original information back! Yes, in theory you can
reverse the profile but then you end up with an image twice removed
from the original.

Hence "corruption".

Now, as I always say, how much of it is visible with our 8-bit eyes -
or whether we even care - is another story, but there's corruption,
all right.

*If* the profile does its job,
it's a better way of getting colors on your screen that match the
original media. Whether you want the colors from the original media is
another question.

Exactly! )

Unfortunately, this is rarely the case and exactly the point why -
unlike (!) monitor or printer profiles - scanner profiles are of very
limited (if any?) use.

Don.

 

[Lorenzo J. Lucchini]

Actually it is! This is why:

When you make your own corrections you can always go back to your
backup/archive and start again.

Once the scanner has *applied* (not assigned!) the profile there's no
way to get the original information back! Yes, in theory you can
reverse the profile but then you end up with an image twice removed
from the original.

Hence "corruption".

Now, as I always say, how much of it is visible with our 8-bit eyes -
or whether we even care - is another story, but there's corruption,
all right.

I realize that you're right in theory, but I'm not sure this is really
an issue in practice.

If you're scanning at 16-bit, well, there will certainly be corruption
due to multiple edits, but it won't have any real impact: you just have
a lot of "spare bits" to work with.

If you're scanning at 8-bit, then what you say holds. However, some or
many scanners (many flatbeds, at least, as I've said) *work at 16-bit
internally* even when outputting 8-bit.
This means that applying the profile *inside* the scanner will probably
be best.

Well, in theory this also means that you should apply *all* of your
edits inside the scanner (actually combining them into one edit), but
that's a bit impractical. Personally, I'd just scan at 16-bit in that
case (i.e. if I know I'm going to heavily edit afterwards, and want good
colors).

[snip]

by LjL
(e-mail address removed)

 

[Don]

I realize that you're right in theory, but I'm not sure this is really
an issue in practice.

It is if one cares for maximum quality. If one doesn't, it's not! ;o)

If you're scanning at 16-bit, well, there will certainly be corruption
due to multiple edits, but it won't have any real impact: you just have
a lot of "spare bits" to work with.

And that's the key! By the time you reduce to 8-bit for
viewing/printing all of that "16-bit corruption" will disappear.

If you're scanning at 8-bit, then what you say holds. However, some or
many scanners (many flatbeds, at least, as I've said) *work at 16-bit
internally* even when outputting 8-bit.

It doesn't matter what internal bit depth is used. If you do
*anything* to the 8-bit image afterwards, image quality will suffer.

On the other hand, if you take this 8-bit output "as is" and do
*nothing* to it, then yes. The question is how realistic is that?

This means that applying the profile *inside* the scanner will probably
be best.

Not if you have to remove the damage the profile has done. Especially
not if that damage has to be removed in an 8-bit file!!

Well, in theory this also means that you should apply *all* of your
edits inside the scanner (actually combining them into one edit), but
that's a bit impractical.

Exactly! (Actually, a *lot* impractical!)

Personally, I'd just scan at 16-bit in that
case (i.e. if I know I'm going to heavily edit afterwards, and want good
colors).

Exactly #2! ;o)

Don.

 

[false_dmitrii]

On Wed, 12 Oct 2005 15:03:00 +0200, "Lorenzo J. Lucchini"


And that's the key! By the time you reduce to 8-bit for
viewing/printing all of that "16-bit corruption" will disappear.


It doesn't matter what internal bit depth is used. If you do
*anything* to the 8-bit image afterwards, image quality will suffer.

On the other hand, if you take this 8-bit output "as is" and do
*nothing* to it, then yes. The question is how realistic is that?


Not if you have to remove the damage the profile has done. Especially
not if that damage has to be removed in an 8-bit file!!

Maybe a different way to look at it is to consider the color shifts the
scanner introduces to the photo as "corruption" or "noise". Bart vdW
was describing in another thread how boosted exposure can decrease
"photon shot noise" with the drawback of additional, but less overall,
dark area noise. In the same sense, using a profile to alter the image
data can restore the original color data at the expense of discarding a
small portion of the bit depth of the scanner output. If the result
matches the original colors better than any subjective adjustment
could, there's a *decrease* in the "corruption" of the original colors
despite the slightly increased corruption of the intermediary scanner
data. In other words, it's an exchange of more one type of "noise" for
less of another...even for 8-bit images, under ordinary conditions.

That's all irrelevant if one considers the scanner data to be the
starting point for all future adjustments rather than an intermediate
point between an accurate transfer of the original image to onscreen or
printed output. You've made it very clear that you want your own
images to come through *exactly* as the scanner sees them at its
optimal hardware settings. Of course, if worried about future output
needs, one can always store the original scan without color adjustment
*and* apply a profile and color space conversion to a copy (provided
one has the right software ). Ultimately, the profiling decisions
all come back to one's original scanning objectives.

false_dmitrii

 

[Lorenzo J. Lucchini]

It is if one cares for maximum quality. If one doesn't, it's not! ;o)

I didn't mean that...

[snip]

If you're scanning at 8-bit, then what you say holds. However, some or
many scanners (many flatbeds, at least, as I've said) *work at 16-bit
internally* even when outputting 8-bit.

It doesn't matter what internal bit depth is used. If you do
*anything* to the 8-bit image afterwards, image quality will suffer.

But then, that's a question of "using 8-bit instead of 16-bit", it has
nothing to do with using a profile or not.

On the other hand, if you take this 8-bit output "as is" and do
*nothing* to it, then yes. The question is how realistic is that?

I don't know. But, as far as the profile is concerned, *you don't do
anything* do the 8-bit output, as the color correction *is done to the
16-bit scan* that the scanner has access to internally.

If we want to discuss 16-bit vs 8-bit, we can, and we have, but that's a
separate issue from that of "can applying a color profile at scan time
cause corruption"?

The answer is: it can't, provided that you *either* work at 16-bit *or*
you work at 8-bit *but the scanner applies the profile internally*
(hence, on its own 16-bit data coming from the sensor).

Not if you have to remove the damage the profile has done. Especially
not if that damage has to be removed in an 8-bit file!!

How can you say it's "damage"?
In theory, a decent color profile should give decent colors.
In practice, there might be many reasons why this need not be true.

But it's unpredictable: with one scan (film type such, density such,
etc) the scan might be closer to "correct" without the color profile
applied, while with another scan (another film type, other densities,
etc) the scan might be closer to "correct" with the color profile applied.

So, your argument might be an argument in favor of scanning at 16-bit
(don't have to "risk" on whether or not to apply a profile), but it
isn't an argument against applying color profiles inside the scanner.

Exactly! (Actually, a *lot* impractical!)

Granted. But even if you *don't* do all the edits inside the scanner
(since it's impractical), you still lose nothing by doing *color
correction alone* inside the scanner.

Rather -- you *might* lose something, but then you might also lose
something by *not* doing it.

I can really only see two solutions to this dilemma: either do
everything inside the scanner (impractical, we know), or scan at 16-bit.

But again, this doesn't have much to do with color profiles, in itself.

[snip]

by LjL
(e-mail address removed)

 

[Don]

Maybe a different way to look at it is to consider the color shifts the
scanner introduces to the photo as "corruption" or "noise".

We have to view it in context. If the image contains a cast in the
same direction as the profile correction than the end result is going
to be twice as bad as if no profile were used. In other words, the
scanner profile is going to amplify the image cast.

And since scanner profile corrections are virtually always non-linear
removing them later is not a trivial operation - never mind the
"corruption" introduced by having to correct the image twice just to
get back to square one!

As I said, if scanner profiles were only tagged, then it would be no
problem. The problem is only when scanner profiles are applied i.e.
the original data is changed.

Bart vdW
was describing in another thread how boosted exposure can decrease
"photon shot noise" with the drawback of additional, but less overall,
dark area noise.

That's different. It's the very reason why I twin scan in order to
eliminate noise in dark areas. However, this opens up the Pandora's
Box of tone response - which is non-linear (!). To then pile on a
scanner profile (also non-linear) just doesn't make any sense in light
of what I just described above. It just complicates things even more
and is a case of diminishing returns.

In the same sense, using a profile to alter the image
data can restore the original color data at the expense of discarding a
small portion of the bit depth of the scanner output. If the result
matches the original colors better than any subjective adjustment
could, there's a *decrease* in the "corruption" of the original colors
despite the slightly increased corruption of the intermediary scanner
data. In other words, it's an exchange of more one type of "noise" for
less of another...even for 8-bit images, under ordinary conditions.

That's a big "if"! And you are not discarding a small portion of bit
depth but a considerable one! Not to mention it's all non-linear and
therefore difficult to remove afterwards should you decide the
inherent scanner bias actually helped the image.

Again, it's up to each person to decide for themselves, but scanner
profile correction is - in many cases - a case of diminishing returns.

I'm not really against scanner profiles as such (and certainly not
against the concept) but unlike monitor or printer profiles which are
both non-destructive and actually help the process, scanner profiles
are of comparatively little use once we consider the whole context
i.e. additional edits down the road.

That's all irrelevant if one considers the scanner data to be the
starting point for all future adjustments rather than an intermediate
point between an accurate transfer of the original image to onscreen or
printed output. You've made it very clear that you want your own
images to come through *exactly* as the scanner sees them at its
optimal hardware settings.

Indeed I have! Repeatedly! ;o) However, what you're talking about is
not in the same league as the *massive* cast Nikons introduce to
Kodachromes! It's primarily that I refer to when I say I want the
scanner to "faithfully" reproduce what it sees. I'd be dancing in the
aisles if Nikons only introduced your average, plain vanilla, scanner
bias to Kodachromes! ;o)

In other cases scanners may introduce assorted casts but we're talking
about a relatively limited (but, I grant you, noticeable) changes, say
on the order of perhaps 10 "clicks" on the 8-bit scale. However,
nothing comes even close to what Nikons do to Kodachromes where blue
clips severely while red is still loitering around 100 on the 8-bit
scale i.e. hasn't even reached the midpoint!

Of course, if worried about future output
needs, one can always store the original scan without color adjustment
*and* apply a profile and color space conversion to a copy (provided
one has the right software ). Ultimately, the profiling decisions
all come back to one's original scanning objectives.

On all of that we absolutely agree!!

Don.

 

[Don]

But then, that's a question of "using 8-bit instead of 16-bit", it has
nothing to do with using a profile or not.

Yes it does if you need to correct this profile "damage" afterwards.

In your scenario above (1st paragraph) the fact that the profile was
applied in 16-bit is irrelevant if all you have to correct it in
afterwards is 8-bit.

I mean, you can't have it both ways. You state that since the profile
is applied in 16-bit *internally* all is well. That's true but only as
long as you don't do anything to the 8-bit output later. And that's
not realistic. Virtually every image will need to be corrected
further. Often in the *opposite* direction of the profile change!

I don't know.

It's not very realistic if one cares for quality. Virtually not a
single image is OK as is i.e. how it comes off the scanner.

But, as far as the profile is concerned, *you don't do
anything* do the 8-bit output, as the color correction *is done to the
16-bit scan* that the scanner has access to internally.

Which is irrelevant if this correction is *wrong*! What then? In your
scenario the only option is to correct the 8-bit output.

If we want to discuss 16-bit vs 8-bit, we can, and we have, but that's a
separate issue from that of "can applying a color profile at scan time
cause corruption"?

It's not a separate issue if your basic premise is based on 16-bit vs
8-bit! You can't have it both ways. If 16-bit is good internally, it's
good externally as well.

The answer is: it can't, provided that you *either* work at 16-bit *or*
you work at 8-bit *but the scanner applies the profile internally*
(hence, on its own 16-bit data coming from the sensor).

Which is all fine, as long you don't touch that image afterwards.

How can you say it's "damage"?

OK, here's an example. Let's say that:

1. The scanner has a blue bias. The profile corrects it by cutting
blue.
2. The image on film has a yellow bias.

Apply the scanner profile and the scan will become even *more* yellow!
That's exactly the *opposite* of what you want! So you have to correct
twice as much. That's damage!

Scanning without a profile in the above example will actually produce
a far better scan.

In theory, a decent color profile should give decent colors.
In practice, there might be many reasons why this need not be true.

And it's the practice we're talking about! It's all fine in theory (no
edits afterwards) but virtually no image is perfect the way it comes
off the scanner.

So, your argument might be an argument in favor of scanning at 16-bit
(don't have to "risk" on whether or not to apply a profile), but it
isn't an argument against applying color profiles inside the scanner.

Yes, it is for the reasons I explained above. Do you really believe
that most images are fine "as is" if the scanner profile is applied?

Granted. But even if you *don't* do all the edits inside the scanner
(since it's impractical), you still lose nothing by doing *color
correction alone* inside the scanner.

You lose *a lot* if the "correction" is in the *opposite* direction of
the final outcome! And, as outlined, even if that is not the case you
still need to edit the image afterwards. Very rarely, if ever (!), is
a scan OK as is when it comes off the scanner - even with the scanner
profile applied!

Don.

 

[Lorenzo J. Lucchini]

On Thu, 13 Oct 2005 20:27:02 +0200, "Lorenzo J. Lucchini"

I mean, you can't have it both ways. You state that since the profile
is applied in 16-bit *internally* all is well. That's true but only as
long as you don't do anything to the 8-bit output later. And that's
not realistic. Virtually every image will need to be corrected
further. Often in the *opposite* direction of the profile change!

"Often"? Why "often"?

You'll have to correct it further just as often as you'd have to correct
further an 8-bit image *that didn't have a profile applied*.

But, really, the chance that you'll need to correct it should be *lower*
if a profile is applied, as a profile's purpose is precisely to give
more faithful colors.
Then if the film colors were wrong for reasons that don't involve the
scanner... tough luck. You'll have to correct after the fact. *But
you'll have to do that regardless of having applied a profile or not!*

Yeah, it *may* happen by chance that the *absense* of color correction
actually compensates for wrong colors in the film.
But the contrary might happen just as easily, more easily actually.

[snip]

OK, here's an example. Let's say that:

1. The scanner has a blue bias. The profile corrects it by cutting
blue.
2. The image on film has a yellow bias.

Apply the scanner profile and the scan will become even *more* yellow!
That's exactly the *opposite* of what you want! So you have to correct
twice as much. That's damage!

Scanning without a profile in the above example will actually produce
a far better scan.

Ok, here is a different example.

1. The scanner has a blue bias. The profile corrects it by cutting blue.
2. The image on film does not have any bias.

Apply the scanner profile and the scan will be just about right. That's
exactly what you want.

If you *did not* apply the profile, then the image would be wrong; so
you'd have to correct it afterwards. That's damage!

Scanning with a profile in the above example will produce a far better scan.


So, it turns out that applying the profile can cause damage (as in your
example); it also turns out that not applying the profile can also cause
damage (as in my example).

How do you decide what to do? Who knows. Looks like an impasse.
But we know we can get out of the impasse by scanning at 16-bit. So, you
see, 16-bit and profile *are two separate issues*.

And it's the practice we're talking about! It's all fine in theory (no
edits afterwards) but virtually no image is perfect the way it comes
off the scanner.

But if a good color profile has been applied, then *the scanned image's
colors should be the same as the original's*.
Which may mean they're still wrong, since they could have been wrong in
the source. But, on average, you'll have done a step forward by applying
the profile.

And in the cases you end up with a step backwards, well... if you did
not apply the profile, you would end up with an equal (but actually
greater) number of such cases.

Yes, it is for the reasons I explained above. Do you really believe
that most images are fine "as is" if the scanner profile is applied?

No. But they'll, on average, even less fine if the scanner profile is
not applied.

You lose *a lot* if the "correction" is in the *opposite* direction of
the final outcome!

Conversely, you gain *a lot* if the "correction" is in the *same* direction.
Or, you lose a lot if the "correction" is in the same direction, but you
don't apply the profile.
Or, finally, you gain a lot if the "correction" is in the *opposite*
direction (again), and you *apply* the profile. Yes, this case exists;
but it's no different from the other cases.

And, as outlined, even if that is not the case you
still need to edit the image afterwards. Very rarely, if ever (!), is
a scan OK as is when it comes off the scanner - even with the scanner
profile applied!

Which only goes to say that you should probably scan at 16-bit, profile
or not.


by LjL
(e-mail address removed)

 

[Lorenzo J. Lucchini]

On 13 Oct 2005 10:45:43 -0700, (e-mail address removed) wrote:

[snip]

Of course, if worried about future output
needs, one can always store the original scan without color adjustment
*and* apply a profile and color space conversion to a copy (provided
one has the right software ). Ultimately, the profiling decisions
all come back to one's original scanning objectives.

On all of that we absolutely agree!!

Do you also agree that, *if you're working at 8-bit* (let's say you are,
then whether it's a good idea is another issue), applying the profile
*internally to the scanner* (which is done at 16-bit) *may* (not will,
but may, let's not discuss the probabilities right here) get you better
results than applying it later (possibly on a copy, doesn't matter)?

What then?

[snip]

by LjL
(e-mail address removed)

 

[Don]

Do you also agree that, *if you're working at 8-bit* (let's say you are,
then whether it's a good idea is another issue), applying the profile
*internally to the scanner* (which is done at 16-bit) *may* (not will,
but may, let's not discuss the probabilities right here) get you better
results than applying it later (possibly on a copy, doesn't matter)?

If you're talking about applying a profile in 16-bit (internally) as
opposed to applying a profile to 8-bit (externally) the answer is:

Not "may" but *will* with one *very important* (!) caveat:

Assuming *nothing* is done to the image afterwards!

And that's a very big assumption which virtually never occurs in
practice.

Don.