[filmscanners] RE: 8 bit versus 16

["LAURIE SOLOMON" <laurie@advancenet.net>]

>> but granting that it does for purposes of argument, it may completely
moot the
>> discussion for you but not for others for a number of reasons that they
are trying to
>> tell you but while you are listening you are not hearing.

>I think you're mistaken about that.

Well, that could very well be and it would not be the first time.

> The discussion is about 8 bit vs 16 bit files and tonal manipulations.
>The ONLY thing I am challenging is the need
>to do high bit tonal manipulations to color image files, period...nothing
>more, nothing less.

Yes the discussion is about that; but I think it is probably about much more
which is implicit in and to the discussion being carried on.  Even if that
is the only thing you are challenging, I think that you have stated it much
too narrowly and absolutely as a universal, while then making practical and
empirical exceptions to this universal rule which in effect transforms the
absolute universal into an empirical or probablistic statement.  You have
already granted that there is or could be situations where there is or might
be a need to do high bit tonal manipulations to color images (i.e.,
restorations of some color images of some subject matter).  I think that the
import or significance of your statements is contingent on the purposes that
the original is being scanned for and the uses the scan is being put to. (I
could probably give illustrative examples; but I will not unless you
insist). Even there, the need may not be an empirically necessary
requirement (e.g., a constitutive need to work in high bit) but one of
comfort and perception (e.g., a preferred need to work in high bit);
however, either way it is a need none the less.

One of the main ambiguities and vageries of this discussion which tends to
introduce both confusion and lack of clarity is if all the references to
high bit or 16-bit or more is to image files of scanner captured and
produced data or to the data itself before it is saved to a file; and if the
reference is to high bit data files, are the references to high bit raw
lineal data files or to high bit raw adjusted data files which do not
contain lineal data across the density range.  A number of scanner software
can output high bit lineal and non-lineal raw data TIFF image files where
the lineal version tneds to be on the dark side while the non-lineal version
has adjusted the gamma of the dark end of the spectrum so as to lighten it
and cause the overall result to look as one might normally expect.  Both
tend to be referred to as high bit raw data files.

I see no difference ( and I would think that others might see no difference)
between establishing setpoints and tonal adjustments during the scan using
the scanner driver and software vis-a-vis high bit hardware generated data
and sending it (outputting it) as low bit data image files to a third party
image editing appplication, a monitor display, or a printer and establishing
setpoints during the scan using the scanner driver and software  vis-a-vis
high bit hardware generated data and sending it (outputting it) as a lineal
high bit raw data image file to a third party image editiing program for
tonal adjustment of the high bit data before converting to low bit.  I think
this is a matter of either preference or the quality of ones scanner
software versus the available third party image editing applications.  I do
see a difference between doing tonal adjustments on an already converted
data image file rather than on the high bit file (assuming the originally
established setpoints are good for the purposes and uses of the scanned
image at hand). First, one might be able to obtain a smoother tonal curve if
you have more bit depth to work with; and second, you will have more data
and possibly finer gradations to work with using high bit than using low
bit.

Moreover, if there is no difference where the tonal corrections or
adjustments are done as long as they are done to raw high bit data before it
gets converted to low bit data, then I fail to see where the disagreement
between you and the others is and it does moot the debate between 16 vrs 8
bit files.  However, I recognize that elsewhere you have said that you are
challenging the need to do high bit tonal adjustments since there is no
difference between 8 bit and 16 bit tonal manipulations, which seems to
stand in contradiction to your position as I read it, and which I can only
read as not being contradictory if I read it as saying that there is not a
difference between 16 and 8 bit tonal manipulations with respect to the
rendering of the file at the output stage of display and printing.

>But there is another reason that scanner
>hardware uses more bits, and it has not a single thing to do with tonal
>manipulation ability.  It is density range.  When you scan, your image only
>takes up some "part" of the overall N bits.  That "part" is larger for
>slides, as they have a higher density range, and lower for negatives.

I understand that; but I am not sure of its relevance to the question of
tonal manipulation.  I would use this as a reason why one would want to do
any tonal manipulations on the high bit data which conprises that "part" of
the density range in the overall high bit density range of N bits.

>Setpoints, yes...but tonal corrections can take place in 8 bit color space
>with no visible degradation over high bit space.

Not the argument or point at issue of those with whom you have been arguing
but my argument and point.  You are talking about no visable degradation but
visability is dependent on the available devices which make display possible
( i.e., monitors and printers) which currently are mainly low 24 bit
constrained devices.  I would argue that for purposes of creating a master
archive file which contained the full spectrum of data possible for
potential use in the future when there may be high bit display devices
available for making the 16 bit image visible doing tonal manipulations in
16 bit may be useful - if not important and valuable - and may show a
difference in visable degradation in the future.

However, your response to my comments while informative and grounds for
discussion was not really responsive to my intended thrust of my comments.
I said:
> I think they all recognize that to apply some Photoshop plugins
> and features or even to print they need to convert the file to 24 bit in
> order to do so and that that should take place after the tonal corrections
> and setpoionts have been established.
Here, I was referring to the use of some Photoshop plugins and features
which will only work in 8 bit and that the application of these features and
plugins should come after all tonal manipulation has been done.  So both set
points and tonal corrections in that case wold need to be done before the
application of the plugins and Photoshop features not after.  In short, they
realize that ultimately the file will wind up as a 24 bit color data file
even if they make tonal adjustments in high bit.

>> There is also the fact that, while
>> one can establish set points and tonal curves that match the optimum
>> hardware capabilities of the given scanner and that these can remain
>> constant...

>But you don't set setpoints and tonal curves to match the scanner, at least
>I don't...nor should anyone, really.  You set them to match the image.

I agree.  I thought you were saying that you did set them to match the
scanner from your statement: "if you get the setpoints and tonal corrections
reasonably close to the scanner driver using high bit...."  The "reasonably
close to the scanner driver using high bit" is what threw me off.

>Setpoints MUST be done to high bit data,

With respect to the term and concept of "setpoints," there is some
confusion.  There is what I will call "native setpoints" which are the two
end points of the density range that the hardware is capable of capturing as
a physical constraint if you will; then there is what I will call "raw
setpoints" which define the practical empirical end points of the density
range that characterizes the original subject that is being scanned and
limits what the scanner could capture even if its native desity range
capablity was great and extended beyond that of the scanned subject
significantly.  In other words, the raw setpoints define the part of the
density range that the scanner as a device can capture occuppied by the
density range of the original scanned subject matter.  In contrast and
conter distinction to the two above, there is also black point and white
points along with the gray point which can be set and manipulated in either
16 bit or 8 bit modes, which are often alluded to as setpoints, which I
would call the "working setpoints" that one defines in accordance with ones
purposes, intent, and preferences to produce a density range or tonal range
and mid point for an image file which meets ones tastes and purposes.  Given
this, I am not sure always what setpoints you are referring to - especially
when you say they need to be established in high bit.  In the latter sense,
setpoints are not more deterministic than tonal or density curves and they
are very much part of the discussion.  I often find that the scanner
captures tonal extremes for many images which are not needed and which if
the two set points are located closer to each other results in a manipulated
but much better tonal curve and density range for the resulting image.

In light of this, it is very possible that you and I could be talking past
each other and that others might also be talking past each other and you
when references are made to "setpoints."

-----Original Message-----
From: filmscanners_owner@halftone.co.uk
[mailto:filmscanners_owner@halftone.co.uk]On Behalf Of Austin Franklin
Sent: Wednesday, September 17, 2003 11:55 PM
To: laurie@advancenet.net
Subject: [filmscanners] RE: 8 bit versus 16


Hi Laurie,

> >But...here's the rub.  If you get the setpoints and tonal corrections
> >reasonably close in the scanner driver, keep in mind, this is all done
> using
> >high bit data...  it's just how scanners work...it completely moots the
> >discussion of 16 vs 8 bit files...as there would be no need to do large
> >tonal moves post scanning.
>
> The other shoe?
>
> First, it assumes that the software being used allows for this;

Yes, I understand that...and have been told that some software doesn't have
decent tools, like a decent setpoint tool, and a decent curve tool, and a
histogram display...

> but granting
> that it does for purposes of argument, it may completely moot the
> discussion
> for you but not for others for a number of reasons that they are trying to
> tell you but while you are listening you are not hearing.

I think you're mistaken about that.  The discussion is about 8 bit vs 16 bit
files and tonal manipulations.  The ONLY thing I am challenging is the need
to do high bit tonal manipulations to color image files, period...nothing
more, nothing less.

> Among
> those is the
> fact that since as you say if you get the setpoints and tonal corrections
> reasonably close to the scanner driver using high bit, you have done
> essentially what they say they are doing just they have choose to do the
> setpoint correction and tonal adjustments to the high bit data in a third
> party editor rather than in the scanner...

The only point of my mentioning that the scanner uses high bit data to do
that was simply to mention it.  But there is another reason that scanner
hardware uses more bits, and it has not a single thing to do with tonal
manipulation ability.  It is density range.  When you scan, your image only
takes up some "part" of the overall N bits.  That "part" is larger for
slides, as they have a higher density range, and lower for negatives.

> ...even if it is possible with the
> scanner software that they are using since they feel more
> comfortable with
> the third party image editor or it is better than that of the scanner
> software.

Not something I have any issue with.

> I think they all recognize that to apply some Photoshop plugins
> and features or even to print they need to convert the file to 24 bit in
> order to do so and that that should take place after the tonal corrections
> and setpoionts have been established.

Setpoints, yes...but tonal corrections can take place in 8 bit color space
with no visible degradation over high bit space.  At least in the hundreds
of images I tried this with...and that's not to say there isn't an image
that may benefit from high bit manipulations, but I've yet to see it...and
no one's able to produce it ;-)

> There is also the fact that, while
> one can establish set points and tonal curves that match the optimum
> hardware capabilities of the given scanner and that these can remain
> constant...

But you don't set setpoints and tonal curves to match the scanner, at least
I don't...nor should anyone, really.  You set them to match the image.  The
images can vary greatly, depending on the film, exposure etc.

> ...for most of us, the subject matter being scanned does not remain
> constant and may require modification of the set points and tonal curves
> from scan to scan or so and is dependent on the original being scanned
> rather than the devices capabilities per se.

I've never said any differently, and that is how I scan as well.

> Such modifications
> may best be
> done after the scan in an image editor where one can actually preview the
> consequences of proposed adjustments on the fly in real time as they are
> being done.

Hum...my scanner software does exactly that...  But, this is really not the
issue at hand.  The issue at hand is only tonal curve manipulation in 8 bit
vs 16 bit...not setpoints.  Setpoints MUST be done to high bit data, and
I've never said any differently.

> In that case the establishment of setpoints and tonal curves
> for the scan should be taken as merely a preliminary raw approximation
> rather than the final product with the main tonal moves being done post
> scanning.

Even if you could get high bit data from a setpointed and tonal curved scan
(which in my experience most scanners either give you
non-setpointed/non-tonal curved high bit data, or setpointed and tonal
curved 8 bit data), I still contend that tonal curve manipulation won't be
any better because you do it with high bit data.

> Second, it very well can be the case that one does not wish to
> replicate the
> exact setpoints and tonal curve of the original that was scanned but to
> deliberately alter or modify the tonal character of the scanned result for
> artistic or other reasons; there would be every reason for doing major
> tonal moves post scanning.

OK...but again, the contention was only tonal curves, not setpoints.  But,
setpoints aren't really an issue either.  Yes, they must be initially
applied to high bit data and MUST be simply because that is how scanners are
designed, but the setpoints should contain all the valid image data, and
therefore allow for anything you mention to be able to be done after the
fact...and therefore only need to be set once.  They are far more
deterministic than tonal curves.

Regards,

Austin

----------------------------------------------------------------------------
------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title
or body


----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body