emulsion side down and histograms

  • Thread starter Thread starter Linda
  • Start date Start date
Linda said:
I took your advice but I'm still having a problem. When I look at my
slides with a Kodak projector or on a light table with a loop they are
nice with bright whites. However all my scans are underexposed. When I
look at the histogram in PS it is all scunched to the left. If I set
the white I have a range of 0 to 78. Am I making sense.

As I said I'm using an Epson 4870 with siverfast SE on a Mac system
9.2.2. I've looked everywhere I can think and other than auto expose
(yuk) I can't figure out how to set the exposure. The scanner has had
very little use. Is this a hardware problem, a software problem or, as
I suspect, my problem?

Yes I do want to preserve as much data as possible. I feel like I'm
just not getting it. Could sure use some more help.

Thanks, Linda (and now I'm of to read, read, read)

In this case the best way to handle a under exposed slide is in the scanner
software.

By adjusting in the scanner software, the scanner will increase the scanning
time to get as much out of the dark slide as possible.

I have several under exposed slides (about 2 stops under) and the resultant
scan to show.
Ask and I can put a few of them on my web site.
 
Linda said:
I took your advice but I'm still having a problem. When I look at my
slides with a Kodak projector or on a light table with a loop they are
nice with bright whites. However all my scans are underexposed. When I
look at the histogram in PS it is all scunched to the left. If I set
the white I have a range of 0 to 78. Am I making sense.

As I said I'm using an Epson 4870 with siverfast SE on a Mac system
9.2.2. I've looked everywhere I can think and other than auto expose
(yuk) I can't figure out how to set the exposure. The scanner has had
very little use. Is this a hardware problem, a software problem or, as
I suspect, my problem?

A couple of things:
- get the latest version of silverfast se. The newer ones have a smarter
auto exposure and may have some additional exposure options.

- get the documentation and demo videos, if you don't already have them.

- In silverfast, look for Histogram and the Gradation Adjustment
(curves) button in the top row on the left panel. Either of those will
allow you to make adjustments to the brightness/contrast of the image.

- In silverfast, have you tried using the "highligh-shadow" dropper to
set your highlights and shadow areas?
 
Linda said:
As I said I'm using an Epson 4870 with siverfast SE on a Mac system
9.2.2. I've looked everywhere I can think and other than auto expose
(yuk) I can't figure out how to set the exposure.

I don't know about the Silverfast interface for these scanners, but with
the Epson scanners TWAIN interface you need to be in full manual mode to
get manual exposure control. When the scan starts hit the "Cancel"
button, which will enable full manual mode. The exposure control is
then available by selecting the tools button.

Don't disparage auto-exposure though. It is completely different form
autoexposure on a camera, for several reasons. Firstly, on a camera the
sensor usually averages over a significant area of the frame to get the
light reading, even for a camera with a spot metering system, while the
scanner uses the data from a couple of pixels in the image. Secondly,
the dynamic range that the camera has to cope with is potentially
infinite (the meter probably has a range of 17-20EV) while for the
scanner the range is already limited by the recording capabilities of
the film. Thirdly, the film response is more 'S' shaped, so that the
response falls off towards the highlights and shadows, which squases the
contrast of the image recorded by the film in the highlights and shadows
compared to the original scene. This means there isn't really a white
or black *point* on the film response, just a region where the response
reduces gradually to zero. On the scanner however, there is a very
definite white point, and the sensor is extremely linear in response ot
to that saturation level. Consequently the scanner autoexposure
algorithm has a very precise white point to target and that will also
guarantee the maximum dynamic range in the final image. Finally, the
algorithm used in the camera is completely different from that in a
scanner. In the camera, the metered measure is used to expose the film
to produce a mid grey, light or dark image depending on the user's
decision to expose normally, for white or shadow. However it does not
expose for white or black because the camera must allow enough headroom
within the film's response to be able to cope with the fact that there
will always be a range of light levels within the patch that is being
metered - so some areas will be brighter and some darker than the meter
reading itself. In the scanner, the autoexposure algorithm can adjust
the exposure so that the white really is white, without risk, because
the measure is made on the lightest few pixels in the prescan.

The bottom line is that autoexposure is scanners is far more precise and
reliable than autoexposure on a camera. Camera makers usually claim
something like 90% of typical images can be accurately exposed
automatically. In practice it depends on your images and could easily
be less than 50%. For scanners, the figure really is higher than 99% of
images that can be accurately exposed automatically. Provided the
preview image is properly framed, I have never found the auto-algorithm
to fail on any scanner I have used.
 
I took your advice but I'm still having a problem. When I look at my
slides with a Kodak projector or on a light table with a loop they are
nice with bright whites. However all my scans are underexposed. When I
look at the histogram in PS it is all scunched to the left. If I set
the white I have a range of 0 to 78. Am I making sense.

Yes. What that usually means is you did not set the exposure properly
when you scanned. Exposure is a hardware setting which can't be
recovered later because there just isn't enough data to do that.

Another possibility is that you set gamma to 1.0 in your scanning
software. If that's the case you need to apply gamma appropriate for
your system in Photoshop (usually 2.2 for Windows and 1.8 for the
Mac).

Also, it may be a case of mismatched profiles. In other words your
scanner profile does not match your PS profile.

And I'm sure there are others, but the most likely culprit is the
exposure.
As I said I'm using an Epson 4870 with siverfast SE on a Mac system
9.2.2. I've looked everywhere I can think and other than auto expose
(yuk) I can't figure out how to set the exposure. The scanner has had
very little use. Is this a hardware problem, a software problem or, as
I suspect, my problem?

Not necessarily. There are all sorts of "gotchas" in scanning. I'm not
familiar with SilverFast so someone else will have to jump in. One way
to test this is to try native Epson software and see if that gives you
a different result.
Yes I do want to preserve as much data as possible. I feel like I'm
just not getting it. Could sure use some more help.

Thanks, Linda (and now I'm of to read, read, read)

Tell me about it...! I waltzed into this groups innocently years ago,
just to ask a few questions and scan my slide collection.

Now - 3 years later - I ended up writing my own scanning program, a
true 16-bit histogram program and an HDR merge. Ugh... I sure didn't
bargain for that when I started!!

If you do want quality, scanning is a road filled with obstacles. I've
done a few things in my life but nothing has been so frustrating and
so full of "ambushes" like scanning. So I sympathize with you fully!

But, if you persevere, there is light at the end of the tunnel! And
it's not always a train coming from the opposite direction! ;o)

Don.
 
Don said:
Not only is that possible but it's a must in many cases if you want to
get the most information out of film in your scans.

So it's not a misconception but a basis for so-called "High Dynamic
Range" images. There are a whole slew of different methods and
procedures to do just that.

My comments were in the context of *a single scan*, and thanks to people
like yourself, I am aware of the different methods to extend dynamic
range by using multiple scans. But all these methods can get complex
quickly if you want the results to be of ultimate quality. Few users
would have the expertise, time and patience to spend three years
developing their own scan programs like you did.

Aside from the online tutorials, a few books do touch on these methods,
BUT they are not specific to scans. I am aware of no scanning books that
go into the kind of details provided by you and Kennedy, etc. in this
ng. On the contrary, the scanning books and online scan tutorials like
Scantips tend to suggest that a scanner's hw is capable of extending a
single scan's dynamic range. They probable contribute a lot to the film
photographers' misconception.

With the rapid advance of digital cameras, film scanners are now much
less popular. It is probably too late for people like you and Kennedy to
come out with a good scanning book. Such a book would have saved me
buckets of blood, sweat and tears over the past three years.
 
However all my scans are underexposed.

P.S. It just occurred to me that may explain why you had problems in
the first place. Namely, correcting the histogram in scanning software
only masks the problem. In other words, the exposure was never correct
but this was "hidden" when scanning software adjusted the histogram.

However, by scanning "raw" and doing the editing in PS later you had a
chance to examine the image in detail before editing which exposed the
exposure problem more clearly.

Don.
 
- In silverfast, look for Histogram and the Gradation Adjustment
(curves) button in the top row on the left panel. Either of those will
allow you to make adjustments to the brightness/contrast of the image.

That's exactly what she's trying to avoid!

Linda wants to scan raw and do the editing in Photoshop afterwards.

Don.
 
My comments were in the context of *a single scan*, and thanks to people
like yourself, I am aware of the different methods to extend dynamic
range by using multiple scans.

Yes, you're right, a single scan does not provide the necessary
dynamic range.

Indeed, in that sense, your film reference is also correct which is
why photographers are forced to perform their own "high dynamic range"
by bracketing exposures and then "merging" the images in the darkroom.
Of course - with film - this is only possible if the subject is
stationary and a tripod can be used to minimize the mismatch between
the two shots.
But all these methods can get complex
quickly if you want the results to be of ultimate quality. Few users
would have the expertise, time and patience to spend three years
developing their own scan programs like you did.

Not only did I spend years on this but also a considerable amount of
nerves and I have the gray hairs to prove it! ;o)

But you're right here as well. Most people don't go through all this
trouble and are (easily?) satisfied with what they initially get when
they start scanning. Which is probably why this is not often addressed
in a lot of detail in various tutorials and other references.

It's a shame, though, because in the end it's not really that
complicated and with the right software the process can be vastly
simplified and streamlined.

Don.
 
Kennedy said:
The main advantage, at least with the Nikon scanner range, is that
NikonScan is a true 16-bit per channel package, whilst Photoshop is
(unless things have changed in CS) only a 15-bit per channel package.

Is there a simple way to test whether a particular program uses 15+1 or
16-bit calculations? Something that could be done without special
tools (such as a verified 16-bit editor or custom software)?

false_dmitrii
 
Linda wrote:

That said I went to the Epson manual (for what that's worth) and read
all your suggestions and decided emulsion side down.

Can't tell from your sentence whether you found the Epson manual info
you were looking for. For anyone interested, the info *is* there, on
page 7. They say base side down, but when I posted the same question
as you sometime last year, emulsion side down sounded like the right
way to go. I don't think Epson has found a way around the laws of
physics yet....

Mr Feinman, I so enjoy your website. Got a bit carried away with the
photos so just started on the tips.

I agree, a nice resource.
I've also decided to scan "raw" and let my partner do the corrections
in photoshop. It's very difficult to correct (edit) someone else's
artwork, especially since we all see different. And as Don suggested I
don't want to lose any data.

If they're your partner's photos, you're probably right to pass on
trying to get the colors just right. Sometimes the original scene will
be very different from what a "sensible" correction creates--only the
original photographer would know. (Nothing wrong with making copies
for your own experiments, of course, as long as you have permission. :)
)

It's important to start with a high-quality image, but always remember
that what the scanner (especially a flatbed) sees is only an
approximation of the original film, and you'll get slightly different
results with almost every scan. You'll always lose or alter some of
the data. :) The ultimate goal is usually to get something very close
to the original with no noticeable defects...a "raw" 48-bit original
will help achieve this by giving you something to go back to if your
adjustments go too far. It's also a very handy foundation for PS
adjustment layers, letting you *always* tweak your corrections without
worry. :) I figure that as long as you're happy with the results of
your careful edits, color management, etc., and you have a copy of the
original scan for days when you become *unhappy*, it's unnecessary to
lose any sleep over all the data that was "destroyed" by the edits--it
was flawed data to begin with. :)

Bit of a blanket statement, but I hope the intent comes through.
I would like to try some scans adjusting the white point. I also see
mention of setting the exposure. I can't figure out yet how to do
either. Suggestions welcome. I did download silverfast docs and will go
through those.

White point adjustment ought to be just as good in Photoshop. It's a
software-only post-scan adjustment, and if you get it wrong in the
scanner software, you'll have less headroom for further adjustments
before the image starts to visibly degrade. It's also harder to find
the "right" point in the small preview window, though not impossible.
Silverfast is much better for adjusting color balance than Epson Scan.
:)

I don't think you can set the 4870's exposure in Epson Scan or
Silverfast, though it obviously has a range of internal exposure
settings. Did someone once write that Vuescan offers the 4870 owner a
choice between 7 or so exposure levels?

false_dmitrii
 
Kennedy said:
[snip]
Is it better to adjust the levels (histogram) with the scanning
software or in Photoshop? I don't want to lose information when
scanning. I've read scantips.com, searched this group and googled. I
have alot of slides of artwork to scan. I don't want the fastest way,
I'd like to do it the best way.
Depends on the capability of the scanner and the modes that you are
using. As a minimum, no matter what scanner and mode, adjust the white
point (for slides) or the black point (for negatives) before scanning.

For scanners like the Nikon Coolscans and Minolta 5400, is adjusting the
white/black points before scanning done by the scanners' hw or sw? If it
is done by hw, are you referring to adjusting the light source' exposure
(the Coolscans' LED intensity and the 5400s' exposure time)? For
scanning slides, I don't believe it is possible to *only* adjust a light
source to get a good white point without losing ground at the black
point, and vice versa. It is kind of like setting an exposure on a film
camera for either highlight details or shadow details, but not both.

The scanner's sw or PS can of course do both. But if adjusting
white/black points is done by the scanners' sw, what is the advantage of
doing it before scanning over doing it in PS after scanning?

<snip>

I believe the SE5400 driver still does some or all of its adjustments
using 8-bit calculations. I remember getting sawtooth histograms with
minimal effort. Can someone confirm or deny this? If so, it would be
a *disadvantage* to adjust color in the SE5400 software instead of in
Photoshop.

false_dmitrii
 
Linda wrote:



Can't tell from your sentence whether you found the Epson manual info
you were looking for. For anyone interested, the info *is* there, on
page 7. They say base side down, but when I posted the same question
as you sometime last year, emulsion side down sounded like the right
way to go. I don't think Epson has found a way around the laws of
physics yet....

<snip>
That'll show you how lame I am. I read that, more than a few times, and
interpeted it as emulsion side down. Reading again I now understand
what they're saying. However I agree with you that emulsion side down
sounds like the right way to go.

This thread has helped me to get a start on comprehending some of the
teminology. I started out knowing nothing. So I guess there is progress
happening here.
If they're your partner's photos, you're probably right to pass on
trying to get the colors just right. Sometimes the original scene will
be very different from what a "sensible" correction creates--only the
original photographer would know. (Nothing wrong with making copies
for your own experiments, of course, as long as you have permission. :)
)

It's important to start with a high-quality image, but always remember
that what the scanner (especially a flatbed) sees is only an
approximation of the original film, and you'll get slightly different
results with almost every scan. You'll always lose or alter some of
the data. :) The ultimate goal is usually to get something very close
to the original with no noticeable defects...a "raw" 48-bit original
will help achieve this by giving you something to go back to if your
adjustments go too far. It's also a very handy foundation for PS
adjustment layers, letting you *always* tweak your corrections without
worry. :) I figure that as long as you're happy with the results of
your careful edits, color management, etc., and you have a copy of the
original scan for days when you become *unhappy*, it's unnecessary to
lose any sleep over all the data that was "destroyed" by the edits--it
was flawed data to begin with. :)

Bit of a blanket statement, but I hope the intent comes through.

Yeh, I think so.
White point adjustment ought to be just as good in Photoshop. It's a
software-only post-scan adjustment, and if you get it wrong in the
scanner software, you'll have less headroom for further adjustments
before the image starts to visibly degrade. It's also harder to find
the "right" point in the small preview window, though not impossible.
Silverfast is much better for adjusting color balance than Epson Scan.
:)

I did some experimenting with this and haven't seen much advantage to
doing it in the scanning software. But as I've said it's all a guess
for me at this point, but interesting and helpful to compare.
I don't think you can set the 4870's exposure in Epson Scan or
Silverfast, though it obviously has a range of internal exposure
settings. Did someone once write that Vuescan offers the 4870 owner a
choice between 7 or so exposure levels?

false_dmitrii

If you can set the exposure in Epson Scan or Siverfast I sure haven't
found where or how. I guess Vuescan if next.

Thanks, Linda
 
SNIP
Is there a simple way to test whether a particular program
uses 15+1 or 16-bit calculations? Something that could be
done without special tools (such as a verified 16-bit editor or
custom software)?

You'd need a true 16-bit application and measure white with it.
Photoshop shows 32768 for white in 16-b/ch mode, but that is mapped
from whatever it was before.

Bart
 
linda said:
That'll show you how lame I am. I read that, more than a few times, and
interpeted it as emulsion side down. Reading again I now understand
what they're saying. However I agree with you that emulsion side down
sounds like the right way to go.

This thread has helped me to get a start on comprehending some of the
teminology. I started out knowing nothing. So I guess there is progress
happening here.
If you can set the exposure in Epson Scan or Siverfast I sure haven't
found where or how. I guess Vuescan if next.

Thanks, Linda

I think you said you have a Epson 4870. Using EPSON SCAN, either standalone
or via TWAIN.

You should have the HTML manual on your CD or it was installed with the
software.

The online HTML manual is here:
http://files.support.epson.com/htmldocs/pr48ph/pr48phrf/index.htm

To make manual adjustments to your scans and to scan film, you have to go
into Professional Mode.
Here is how to go to Pro Mode:
http://files.support.epson.com/htmldocs/pr48ph/pr48phrf/howto_1.htm

When you enter the Professional Mode, Notice that there is a Scroll Bar on
the Right.
When you scroll down, you with see more parts of the Scan adjustments.

Scanning Films: All of the adjustments are here.
http://files.support.epson.com/htmldocs/pr48ph/pr48phrf/howto_2.htm#for films b

To make exposure adjustments.
Scroll or Find on the page at the above link. (about mid way down) The
following sentence.

"When grains appear on the scanned image, select the Grain Reduction check
box and then click SCAN again to reduce them. Selecting a higher Grain
Reduction level makes the scanned image softer in texture. "

Just below that sentence is a picture with Grain Reduction marked in Red.
That is not what I want you to see, look at the picture, where it says
"Adjustments", that is where the adjustments to the histogram, white point
and black point and color adjustments are done.

There are four Icons, each Icon is one or the other of the full range of
adjustments that can be done to scans with the Epson 4870.

The manual also says in another part:

The image is too dark
Try one or more of these solutions:


Check your software's Brightness setting using the Image Adjustment dialog
box in the Home Mode or the Histogram Adjustment dialog box and the Image
Adjustment dialog box in the Professional Mode.

Check the brightness and contrast settings of your computer display.

The above from the page:
http://files.support.epson.com/html...-hatch patterns appear in the scanned image b

Epson does not have a clear manual and there is no search ability.
You can search using Windows Search if the HTML manual is on your computer.
 
Don said:
Not only did I spend years on this but also a considerable amount of
nerves and I have the gray hairs to prove it! ;o)
Pah - shirker!! You still have hair?? ;-)
 
Is there a simple way to test whether a particular program uses 15+1 or
16-bit calculations? Something that could be done without special
tools (such as a verified 16-bit editor or custom software)?
The trick is to turn the limited precision into a spatial error that can
be visually assessed. The way I tested it was to create a tiff file
with a 16-bit ramp in it. This is fairly straight forward if you learn
the tiff specification, and something as small as a 256x256 pixel image
will create a grey ramp, or 256x768 pixels a ramp in each r,g,b channel.
Then just import that file into the application, implement a couple of
transforms - such as rotating the image back and forth by 90deg - and
save the result. By examining the data of the saved file in a hex
editor, or by writing some software to compare it with what it should
be, the precision (or lack of it) in the program can readily be seen.

For example. If you implement a gamma compensation on a grey ramp in
Photoshop, odd and even pairs of pixels have the same level, whilst the
same gamma applied in NikonScan results in a smooth curve incrementing
at each pixel.

Another method I have used in the past is to apply a fixed gain in
integer powers of 2 and then examine the histogram and count the levels
that exist. For example, say you start with a true 16-bit per channel
image (which implies scanning with unity gamma, otherwise you have lower
resolution in the shadows in any case). Then go into Levels in PS or
the equivalent in your application under test and set the lower point to
127 and the upper to 129, this will apply a gain of x128 to the image,
biased about the mid range. The result, if a true 16-bit application,
should have 512 levels in the image, is 15bit+1 it will have only 257
levels. Applying a relatively small amount of additional gain using the
levels control will show this in the histogram. For example, taking the
lower level to 64 and the upper level to 191 should apply a further gain
of 2x, so there should be 256 levels in the final result if it is a true
16-bit application. If you do this with Photoshop using the 256x256
linear ramp I mentioned above, which starts with all 65536 possible
levels and view the central 256 pixels of the ramp, you find the
histogram has 128 levels, not the 256 expected of a 16-bit application.

A linear ramp is quite a nice tool for reverse engineering a variety of
processes because you can see what each pixel transforms to and then
work out what function is required to achieve that for all of the
pixels.
 
Pah - shirker!! You still have hair?? ;-)

ROTFL! ;o) I know... Quite amazing after what scanning has put me
through!

Actually, the hair's been down to my waste since my teens. I chose my
maternal grandfather well! ;o)

Don.
 
Is there a simple way to test whether a particular program uses 15+1 or
16-bit calculations? Something that could be done without special
tools (such as a verified 16-bit editor or custom software)?

The simplest way is probably to get a true 16-bit histogram program.

I'm not aware of one, so I wrote my own. It's quite revealing to
examine a histogram of a PS edit where every other bin is empty! Half
your data out the window, just like that... :-(

In general, I found the 16-bit histogram quite an eye opener. All
sorts of things pop-out when put under a 16-bit histogram microscope!

Don.
 
I guess Vuescan if next.

That's a touchy subject around here - for some... ;o)

If you're after quality I personally wouldn't touch VueScan with the
proverbial 10-foot pole. It just far too buggy and unreliable. It may
be OK for casual use to scan a tiny web graphic, but that's about it.

Some consider VueScan a "poor person's SilverFast". Unable to afford
the pricey SF they settle for VS. It's free to try so you can make up
your own mind (although some features like raw scan are disabled) but
once you've paid, no refunds (duped victims often post complaints).

Having said that, there are some dedicated, but touchy, VueScan "fans"
around here who - judging by past history - will now unleash their
insults... ;o)

Don.
 
Back
Top