Re: 35mm slides scanner clarification questions

Today Kennedy McEwen commented courteously on the subject at
hand

It is just the same really - experience tells you what
operations are likely to create discontinuities in the
visible image when you start with 8-bits and those which
won't.

I assume that you already have some 8-bit images from your
comments about using PSP - so you must already have some
experience of whether the processes that you typically use
produce visible tonal discontinuities or not. If they do,
then you will find that 16-bits per channel will be less
prone to the problem - that's all there is to it.

Kennedy, you have no way of knowing, but I've been doing
raster graphics since about 1990 when I bought my first hand-
scanner. I originally used Aldus PhotoStyler, then PhotoShop
5.0 LE, then switched to PSP, starting with v3, then 5, then
7, 8, and 9. X has about 2/3 of its functions in 48-bit, but I
don't like what Corel did to my fav app and haven't upgraded.
I would if I ultimately understand better how to at least
start in 48-bit, and after I've at least started up the
learning curve on RAW.

16-bits per channel just provides a safety factor for
editing, not a perceptible difference in the image, whether
that is on the 8-bit per channel display, an 8-bit per
channel printer or even a 16-bit per channel printer if you
even managed to find one.

Again, I understand the theory, but have no clue how to put it
into practice. I have 16 years of experience, but am not at
all an expert. Hell, I only use about 1/3 of PSP 9's
functionality! If I can ever amass a "critical mass" of
information on 48-bit and RAW, I would start learning
immediately, somehow.

The part I still cannon comprehend is /not/ why 48-bit is
superior in the first place, but how one can /see/ the
difference on an 24-bit.

You won't see *any* difference between a 24-bit colour
image and a 48-bit one. 24-bit colour range already
exceeds the range that you can see, so 48-bits won't make
any difference.

What 48-bit colour does is to permit you to process the
image (typically, but not only, high contrast increases and
severe level and curves adjustments) through PS (or any
other 48-bit compatible software, such as PSP X) without
introducing tonal discontinuities in your image.

I talked about PSP X above, which I don't have yet. The /very/
little understanding of what you and others are talking about
is that I have to learned to use non-visual analytical tools -
I don't know which - and lots of experience as you've been
trying to get my feelble, over rigid brain to accept. I do not
at all dispute you or anyone, I just don't comprehend it, yet.
Eventually, I will, and I'm sure I'll wonder how the hell I
got along without it.

Maybe what I will need is to find some of my old slides that
just cannot be adequately saved with 24-bit software...

For example, say you make a scan on 24-bit colour of a
slide of a black cat in a snow field. Assuming that the
original slide was exposed to show detail of the cat's
black fur then the snow white background might be pure
white. The scanner will usually adjust its exposure to
ensure that the snow just fails to saturate the 8-bit
range, perhaps producing a level around 250 or so.
Meanwhile the black cat could be buried in the shadows -
its entire range being perhaps only 0-30, even though that
is the main subject you are interested. You could
(probably should, but it depends on whether there are other
background details you want to preserve) adjust the scanner
exposure, sometimes (as in the case of Nikon) called
"analogue gain". However, if this was a sample from a
batch it could have been scanned while you were away from
the machine on automatic settings.

Again, I am not taking you on. I understand dynamic range and
pathological cases like your example. In my case, it is often
noise in the shadows and blown-out highlights that mess up my
car pictures.

A post scanning solution would be to adjust the levels in
the image to lighten the cat significantly, bringing up all
of the crushed detail in the shadows, perhaps taking the
cat's range from 0-30 up to 0-150, a contrast gain of x5.
Now, you still have the same 50 levels of data, but they
are each spaced 5 levels apart and you probably will see
some posterisation effect in the texture of the cat.
Instead of a smooth range of tones, you will see discrete
tonal levels - the image looks like it has been made by
silk screen printing, or posterised.

To do what you're describing, though, assumes that there is
any detail in the original you're scanning from. If it really
is an overexposed picture of a black cat on snow, there ain't
any snow flakes left, just miles of pure white.

You could have avoided this effect if the original scan had
been made with 48-bit colour. Now, while the cat occupies
0-30 in the 8 bit scale of the displayed image, there are
really 30x256 discrete levels in the source data (actually
it will be quite a bit less in this example, for reasons
its probably not worth going into at the moment). So now,
when you apply the 5x contrast stretch you get a range of
0-150 on the 8-bit scale of the final image with all of the
levels occupied, no gaps between them and a continuous
smooth range of tones. Effectively you are bringing out
some of those levels in the 48-bit colour source that were
buried *between* each of the levels in the 8-bit displayed
image.

Yu're still losing me, I'm afraid. I understand the basics,
but I know that a good number of my old slides lack detail in
both shadows and highlights, and white is white and black is
black. Or, am I still being so rigid I'm missing the big
picture, here?

Yes, scanners can do better on input, and printers better
on output, so I'm trying to get my arms around how to try
to approach this without having to run test prints on a
printer I don't have in the first place.

You adjust the image in your editor until you are happy
with the way it looks. Assuming a 48-bit colour original
you either print that directly, in which case the printer
driver will reduce it to 24-bit colour transparently to
you, or you reduce it to 24-bit colour yourself and print
it, transfer it to your print shop or whatever output
method you want.

48-bit colour image files are created to provide a safety
factor so that edits can be made without introducing
unwanted artefacts. There is nothing extra to see in them
at all - until you make those edits. Even 24-bit images
have more colour range than you can see, so relatively
severe tonal edits can be performed on 24-bit colour
without any perceived posterisation.

One final issue on this though. Just because you have an
8-bit image from your 16-bit scanner, it doesn't mean that
the scanner is discarding the least significant 8 and only
using the 8 most significant bits. That 8-bit image will
usually be gamma compensated to *both* correct for the
gamma of the monitor and distribute the 8 levels evenly
throughout the range of shades that you can see. To
implement that gamma compensation without introducing any
gaps in the histogram, the scanner driver uses all 16-bits
of the ADC signal range. Similarly, when the scanner
produces a 48-bit colour image it also applies gamma
compensation to that, and so many of the levels available
in the shadow region are unoccupied. This is why, when I
was talking above about number of levels occupied by the
cat in the 48-bit image, I commented that there would
actually be less than 30x256 in total expected.

I under gamma in prinicple, too, but have never learned how to
use it effectively. That's my bad, and not at all a limitation
of the techology I am using - or what I could be using - for
input or post-processing.

The rub for me wrt to whatever scanner I ultimately buy is
that I don't have the time nor skill at this point in time to
do what you're describing, so as of yet, I have no
inclination. But, since necessity is the motherhood of
invention, I may well find that I must quickly ramp up once I
get started. I've estimated that my project will take me a
year or more, no matter how many I can scan at a time,
because, I cannot do it physically with my current non-life
threatening illnesses. For an easy example, I've got maybe
4,000 digitals of cars at shows and museums in just the last
year alone that I just cannot get to.

Also, I am a stickler for having good meta data to describe
the picture via its file name, so my "rigitity" prevents me
from just saving back to IMG_0001 to IMG_4000. I can't find
pictures by searching, I'd have to rely on a thumbnail viewer.
And, if you know about car picture collectors or even more
advanced vacation photogs, they like to have folders full of
pictures under some logical arrangement.

I hope this explains why you are needlessly concerned about
how you will cope with a problem that you will never be
faced with. You don't need to see anything that is hidden
in the 48-bit colour image. The print will be just the
same as if you printed a 24-bit image - no additional test
prints necessary.

It is very doubtful I'll print anything much in future, and
certainly nothing to a custom lab where all this stuff would
be helpful. My main intent is to set up picture "albums" in my
fav thumbnail viewer, Jasc's now defunct Media Center Plus
3.10.

The only difference between the 24-bit
and 48-bit colour range is just a safety factor for edits.

This last paragraph is exactly what I've been searching for a
way to explain to you. I don't /think/ I have a need, so, yes,
I am "needlessly concerned" for nothing. However, at my
pension funding level, if I'm going to stroke $1,600 on a good
scanner, I certainly should try to heed and learn from your
experience and advice to the extent I can, which again, is why
I'm carefully reading, commenting, asking questions, and
saving everything for later, for the day when I do have the
time and inclination.

Thank you again, Kennedy, for your patience and great detail
in explaining this stuff to a scanning and 48-bit newbie.

--
ATM, aka Jerry
.

Relevant Pages

  • Re: 35mm slides scanner clarification questions
    ... histogram eventually appear as posterisation of the image. ... The scanner will usually adjust its exposure to ensure that the snow just fails to saturate the 8-bit range, perhaps producing a level around 250 or so. ... Meanwhile the black cat could be buried in the shadows - its entire range being perhaps only 0-30, even though that is the main subject you are interested. ...
    (rec.photo.digital)
  • Re: Problems pasting TIF image into Notebook document.
    ... It's possible that your scanner software includes some image editing utility ... It could be that the TIFF is greater than 8-bit and/or CMYK color mode, ... If I resize the picture smaller the same proportion of blank band ... The only way I can reduce the width of the band is to omit the right ...
    (microsoft.public.mac.office.word)
  • Re: Problems pasting TIF image into Notebook document.
    ... Scanner settings can easily determine the usability of an ... If I resize the picture smaller the same proportion of blank band ... The only way I can reduce the width of the band is to omit the right ... When I copy a picture in TIF format into a Word Notebook document a ...
    (microsoft.public.mac.office.word)
  • Re: Problems pasting TIF image into Notebook document.
    ... Scanner settings can easily determine the usability of an image. ... It could be that the TIFF is greater than 8-bit and/or CMYK color mode, ... If I resize the picture smaller the same proportion of blank band ... The only way I can reduce the width of the band is to omit the right ...
    (microsoft.public.mac.office.word)
  • Re: Problems pasting TIF image into Notebook document.
    ... If Word can't see the scanner, it does need to be re-installed ... It could be that the TIFF is greater than 8-bit and/or CMYK color mode, ... If I resize the picture smaller the same proportion of blank band ... The only way I can reduce the width of the band is to omit the right ...
    (microsoft.public.mac.office.word)