Scanner exposure settings when gang scanning multiple photos

[skylark]

This topic deals with getting the optimum exposure for EACH photo when
scanning multiple photos at once on flatbed scanners.

First off, do typical (under $200) flatbed scanners adjust the physical
exposure time of the CCD based on the preview scan? I'm referring to
the hardware not software.

If so, then it would seem that if photos are scanned individually, the
optimum dynamic range for each photo could be captured. In contrast,
if multiple photos are scanned in one pass, then wouldn't the exposure
time have to be compromised if one photo was really dark and another
photo was really light?

What about scanning with no preview scan and with no auto corrections
set in the scanner? When I do this using Vuescan in raw mode, after I
crop the photos from the scan and look at the histogram in Photoshop,
the data is only in the middle portion with wide unused borders on the
left and right sides. Optimally I expected to see the histogram fill
the window with no clipping. If I set an auto setting in Vuescan to
get a histogram that fills the window, is it just doing that in
software which would be equivalent to doing the same thing using
Photoshop?

Finally, what if I gang scan multiple photos using my Canon 8400F
software in normal scan mode (not using the scanner's auto crop and
straighten). Then do the auto crop and straighten using Photoshop
Elements 4? Will I be losing dynamic range if there are dark and light
photos in the single scan? I want to do this to scan at higher
resolutions than the Canon auto crop and straighten software's 300 dpi.

Thanks for any insights,
Skylark

 

[Roger S.]

I'm sorry, what are you scanning? If negatives, one exposure should
work for a given type of film. If slides you may benefit from
increasing exposure for dense slides. If prints I don't know but would
guess one exposure would work.

"What about scanning with no preview scan and with no auto corrections
set in the scanner? When I do this using Vuescan in raw mode, after I
crop the photos from the scan and look at the histogram in Photoshop,
the data is only in the middle portion with wide unused borders on the
left and right sides. Optimally I expected to see the histogram fill
the window with no clipping. If I set an auto setting in Vuescan to
get a histogram that fills the window, is it just doing that in
software which would be equivalent to doing the same thing using
Photoshop? "

Yep. Make sure you have some headroom on both sides for editing in
Photoshop and output a 16 bit per channel (48 bit) file to work on
later. Having Vuescan expand the histogram isn't all that helpful
(although I do recommend letting vuescan convert from gamma 1.0 to 2.2-
I have not found a use for linear gamma raw files).

 

[skylark]

I'm sorry, what are you scanning?

Photo prints on a flatbed scanner.

Yep. Make sure you have some headroom on both sides for editing in
Photoshop and output a 16 bit per channel (48 bit) file to work on
later. Having Vuescan expand the histogram isn't all that helpful
(although I do recommend letting vuescan convert from gamma 1.0 to 2.2-
I have not found a use for linear gamma raw files).

There is, I think, excessive headroom on both sides. I would estimate
about 20% on each side totaling about 40% wasted dynamic range. I'm
wondering if scanning photos one at a time using a preview scan for
each would result in more dynamic range captured.

Skylark

 

[Roger S.]

There is, I think, excessive headroom on both sides. I would estimate

about 20% on each side totaling about 40% wasted dynamic range. I'm
wondering if scanning photos one at a time using a preview scan for
each would result in more dynamic range captured.

No, dynamic range is the difference between Dmax and Dmin. If there is
extra room on both sides it means you have captured the entire dynamic
range of the print with room to spare. Filling the histogram just
means clipping the excess data. If you scanned each seperately in your
scanner program, the scanner software could do this for you.

Of course Photoshop can do at least as good a job. This should help
you get started:
http://www.creativepro.com/story/feature/17164.html?origin=story

Roger

 

[skylark]

No, dynamic range is the difference between Dmax and Dmin. If there is
extra room on both sides it means you have captured the entire dynamic
range of the print with room to spare. Filling the histogram just
means clipping the excess data. If you scanned each seperately in your
scanner program, the scanner software could do this for you.

Of course Photoshop can do at least as good a job. This should help
you get started:
http://www.creativepro.com/story/feature/17164.html?origin=story

Roger

Thanks for the nice link. Sorry to persist but what I don't understand
is this:

Exposure goes from 0 to 255. So if the scanned image only takes up 60%
of the histogram, doesn't that mean that black is at 51 and white is at
204? If so, then it seems that all of the exposure is compressed
within a total range of 0-153 (255 * 60%) instead of the full 255
range.

So my basic question is is this:

1. I scan without doing a preview scan but the histogram is confined to
60% of the histogram window to insure there is no clipping.

2. I scan doing a preview scan first, then adjust the exposure setting
so that the histogram just fills the entire histogram window and then
do the actual scan.

Would the second scan be better than the first one or would they be
identical? If Ed Hamrick is online, would appreciate your thoughts on
this using VueScan.
Thanks,
Skylark

 

[Barry Watzman]

The results are not the same and I'd argue that choice 2. is higher
quality. Adjusting the exposure to expand the dynamic range (so that an
image that was originally only using 60% of the dynamic range) does not
imply that there is clipping (if you have not excessively adjusted the
exposure on either end).

By the way, to add to your post a bit, the fact that the scanned image
uses only 60% of the available dynamic range doesn't tell you where that
60% lies. 60% of 255 is 154 units; but it could be 0 to 153, it could
be 102 to 255, or it could be any set of numbers from 0 to 255 that are
153 apart. You normally don't want to clip black or white, but the
exact expansion will vary from original to original (and, sometimes, you
might want to clip, for example if scanning a printed page (black
letters on a semi-white page), clipping can enhance the image by making
defects (old gray or yellow pages that are no longer truly white, or
creases in pages that were folded) disappear.

 

[skylark]

The results are not the same and I'd argue that choice 2. is higher
quality. Adjusting the exposure to expand the dynamic range (so that an
image that was originally only using 60% of the dynamic range) does not
imply that there is clipping (if you have not excessively adjusted the
exposure on either end).

Yes, I'm wondering if expanding the histogram in Photoshop to fill the
histogram window would result in banding in graduatations like the sky
or shadows. Where as choice 2 would have smoother transitions in the
sky and shadows.

By the way, to add to your post a bit, the fact that the scanned image
uses only 60% of the available dynamic range doesn't tell you where that
60% lies. 60% of 255 is 154 units; but it could be 0 to 153, it could
be 102 to 255, or it could be any set of numbers from 0 to 255 that are
153 apart.

My multiple photo (prints) scans using Vuescan are with no auto
corrections and saving to a raw file. Then do the preview scan on the
raw file and crop the individual photos. When the photos are views in
Photoshop levels, the histograms show the data portion about in the
middle with about 15% to 20% blank on both sides. The data is not
skewed toward zero or to 255. I'm thinking that I'm losing quality
doing multiple photos scans using Vuescan this way.

Skylark

 

[David J. Littleboy]

Yes, I'm wondering if expanding the histogram in Photoshop to fill the
histogram window would result in banding in graduatations like the sky
or shadows. Where as choice 2 would have smoother transitions in the
sky and shadows.

The "obvious" thing to do is to scan in 16-bit mode.

However, if you remember that dynamic range is (max signal)/(noise floor)
(or the log thereof), then you would realize that the question is a tad
harder. You need to ask what is the largest noise component and what
workflow minimizes that component.

In scans, it looks to me that the grain noise is the largest noise component
(it certainly is in sky areas in negative films), so what you need to ask is
whether adjusting the analog gain at scan time results in less grain noise
than stretching the histogram after the fact. Since stretching after the
fact will exacerbate grain noise by the amount of the stretch, it's possible
that adjusting the gain is a better way. Maybe.

My _guess_ is that there wouldn't be much difference, but that's just a
guess. But every time I mention that grain noise is a problem on a scanner
list, I get flamed. It's the 800 lb unmentionable gorilla...

David J. Littleboy
Tokyo, Japan

 

[Barry Watzman]

If you scan multiple photos on the glass at once, then the scan exposure
parameters can (at best) only be optimized for one of them, and might
well be suboptimal for all of them. While scanning in this manner might
save time, it makes it impossible to get the absolute best quality for
all of the prints on the glass.

 

[Barry Watzman]

My comments relate to scanning prints, not film of any kind, and
therefore as a practical matter there are no grain issues, since the
maximum scan resolution likely to be used for a print is probably 600dpi.

 

[Roger S.]

David's right- scan in 16 bit mode so it won't matter if you do
post-exposure adjustements in Photoshop or the scanner driver. I don't
think you have per channel analog gain with a flatbed scanner and think
the "grain" noise advice doesn't apply to prints.

Increasing or decreasing exposure will move the data left and right but
I don't see how it would expand it. Is your image data currently
squashed in the shadows right now? If not, expanding the image to fill
the histogram is just what your driver is doing in software with an
autolevels-type function and Photoshop can do this at least as well.

Generally with digital capture devices a good goal is to expose to the
right and try to get above the hardware noise floor. Do you see
scanner noise in dark areas of your print?

Just today I have been running tests to see if increasing exposure on
my negatives (which don't come close to filling the scanner historgram)
reduces sky noise and it seems to yield some improvements. Expose to
the right and you'll be all right.

 

[Roger S.]

You normally don't want to clip black or white, but the

exact expansion will vary from original to original (and, sometimes, you might want to clip, for example if
scanning a printed page (black letters on a semi-white page), clipping can enhance the image by making
defects (old gray or yellow pages that are no longer truly white, or
creases in pages that were folded) disappear.

Which is a powerful argument for capturing as much good data as
possible at scan time and then doing the adjustments in Photoshop.

 

[David J. Littleboy]

My comments relate to scanning prints,

Oops. Sorry for the noise.

David J. Littleboy
Tokyo, Japan

 

[Barry Watzman]

Re: "Increasing or decreasing exposure will move the data left and right
but I don't see how it would expand it."

In HP Precision Scan Pro, you have separate adjustments for brightness
and contrast as well as gamma (HP calls them highlights, shadows and
midtones). You can expand the histogram to go from 0 to 255 without
clipping; I've never used a 16-bit readout (which is available, as are
10-bit and 12-bit); I was under the impression that it didn't change the
output anyway, only the display. There are two separate pushbuttons to
show if/where you are clipping on either highlights or shadows. I have
never found the "auto" function to work to my liking however, but
manually it allows tremendous improvement in the quality of the output.
I see no noise at all, never have in any of my material.

Most of the photos that I scan are old ... REALLY old (50+ years) and
are therefore monochrome. Most, but not all.

[Scanner is an HP 5470C]

 

[skylark]

If you scan multiple photos on the glass at once, then the scan exposure
parameters can (at best) only be optimized for one of them, and might
well be suboptimal for all of them. While scanning in this manner might
save time, it makes it impossible to get the absolute best quality for
all of the prints on the glass.

I guess part of what I don't really understand is for a given set of
photos on the glass, whether the data captured from the scanner's CCD
is always the same and all exposure adjustments are done by the
scanner's firmware after the actual scan? Or does the scanner actually
adjust the exposure time of the CCD hardware based on the preview scan
settings?

With regard to Vuescan's raw mode scanning multiple photos. Suppose
the following steps:

1. Scan multiple photos (prints) to a raw file.
2. Do a preview scan from the raw file.
3. Adjust the marching ants around one photo only.
4. Scan that photo from the raw file.

Would that be identical to scanning that same photo by itself directly
from the scanner as far as the dynamic range captured goes?

It's hard to explain what I'm really after but I hope this makes some
sense,
Skylark

 

[Roger S.]

I guess part of what I don't really understand is for a given set of
photos on the glass, whether the data captured from the scanner's CCD
is always the same and all exposure adjustments are done by the
scanner's firmware after the actual scan? Or does the scanner actually
adjust the exposure time of the CCD hardware based on the preview scan
settings?

I think your scanner software is unclear about what it's actually
doing. While I don't have much experience scanning prints, I've been
using Vuescan a while. Here's an experiment you can do, assuming
Vuescan scans prints similarly to slides and negatives.

Enable the histogram tab and set it to RAW. In Vuescan, pick one
print, hit preview, correct the cropping if necessary, and then hit
lock exposure. Make a note of the recommended default exposure value
and hit preview again to do a preview at the locked exposure value.
Look at the data and see if you have a lot of headroom to the right or
not. If you have a lot of headroom, consider bumping up the exposure
and hitting preview again. Do this until you go too far and you lose
highlight data. Note your settings and what you were scanning.

To see if this makes any practical difference, do a full resolution
scan at the nominal exposure level and another at the higher exposure
level and see if there is any difference in noise or other artifacts.
Zoom into dark shadow areas.

This is pretty time intensive but you shouldn't have to do this for
each print you scan- I'm guessing that similar media behave similarly
so you can reuse settings (say for newsprint versus B&W prints). You
can test this hypothesis by previewing other types of media and seeing
what the recommended exposure is and what their histograms look like.

Please report back your findings so others can learn from it.

With regard to Vuescan's raw mode scanning multiple photos. Suppose
the following steps:

1. Scan multiple photos (prints) to a raw file.
2. Do a preview scan from the raw file.
3. Adjust the marching ants around one photo only.
4. Scan that photo from the raw file.

Would that be identical to scanning that same photo by itself directly
from the scanner as far as the dynamic range captured goes?

No. Dynamic range is determined by exposure, and it would be a one
size fits all exposure followed by custom software processing on each
file if I'm understanding you correctly. Software changes just stretch
the data- they don't increase dynamic range. You will have more than
one print on the glass at a given time, correct?

What is identical to this workflow is that you could output a file with
color balance set to "none" and then do edits in Photoshop. In PS you
could break each picture up into its own layer, or create a selection
box around each one and do adjustments just on the selection, turn them
into their own files, etc.

It's hard to explain what I'm really after but I hope this makes some sense,

Feels like we're going in circles, but try these tests and you'll be
more confident in your workflow...