2 easy exposure questions

[false_dmitrii]

1: Why does Epson Scan's auto exposure always limit the output levels
to 10-200, rather than 0-256? It seems too extreme to me and is one
of the main reasons I dislike Epson Scan. Auto exposure also sets the
gamma/curve for each channel separately, which seems to cause a bad
color shift if I expand the output levels back to their full range.
In my inexperience I haven't come across any recommendation to limit
output levels; is there a compelling reason?

2: Is it best to set exposure levels close to the desired final
output when scanning a RAW file for future use in Vuescan or
Silverfast? Or does the greater amount of image data offer great
leeway in adjusting clipping points and gamma, so that relatively
"empty" areas at the start and end of the histogram can nonetheless be
preserved in the RAW file in case they prove relevant later?

Thanks,
false_dmitrii

 

[degrub]

1) can't comment on Epson in particular. Setting the curves individually
sounds like they are trying to remove any color casts. Limiting the
output levels makes sense to me if the image doesn't have the full
dynamic range available and the software is trying to avoid posterizing
or they are trying to capture as much information as possible without
distorting what it perceives as neutral colors. The whitest white in the
image may not be 250,250,250 to the scanner.

2) Usually, you can change the exposure to fill the histogram levels
pretty close to 0-255. A lot of people restrict the range to somewhere
around 245 to 250 levels. Some images won't have data values at the
extremes . Look out for clipping, where , when you look at the
histogram, there is a "pile" of pixels at either end. Changing the
exposure will help if your scanner hardware is capable.

Frank

 

[false_dmitrii]

1) can't comment on Epson in particular. Setting the curves individually
sounds like they are trying to remove any color casts. Limiting the

That seems to be the effect. The problem is that the varying
graypoint values make it more difficult and time-consuming to tinker
with the auto-exposure values. And there's no way to set
auto-exposure options.

output levels makes sense to me if the image doesn't have the full
dynamic range available and the software is trying to avoid posterizing
or they are trying to capture as much information as possible without
distorting what it perceives as neutral colors. The whitest white in the
image may not be 250,250,250 to the scanner.

Hmm, that makes sense. I checked again and found that the "200" upper
limit is for negatives, whereas prints limit white output to "245",
which was the value I originally associated with auto-exposure. Since
the values don't seem to vary, maybe they're hard-coded for particular
media types.

I had the impression that it was generally a good policy to capture as
much image data as possible when scanning images, including the
fullest possible contrast range, unless the goal is to minimize or
eliminate postscanning adjustments. A 16-bit/channel application can
then apply the output limits if necessary, no?

Is it unusual to scan negatives, with their inherent low contrast
range, using 245 or more output levels? For example, in a photo of a
tortoise, there was a large flash reflection on the tortoise's shell.
The 1-hour photo shop's print left a large "blowout" in this region,
and Epson Scan's auto-exposure did the same. Widening the output
lightness range and setting the color channels manually revealed
considerably more shell within the blowout zone. Similarly, Epson's
settings frequently produce a bright, almost-white sky in images where
more subtlety is present. Am I risking degradation of the image
colors by expanding the negatives' output levels past 200?

To put it another way, it seems to me that using close to the full
lightness range offers the best chance to preserve more of the full
range of the image's shadows and highlights as they would be perceived
in the real world, whereas limiting output levels to 200 means that
more of the perceived shadow and highlight detail must be sacrificed.

2) Usually, you can change the exposure to fill the histogram levels
pretty close to 0-255. A lot of people restrict the range to somewhere
around 245 to 250 levels. Some images won't have data values at the
extremes . Look out for clipping, where , when you look at the
histogram, there is a "pile" of pixels at either end. Changing the
exposure will help if your scanner hardware is capable.

My thinking was that with a 16-bit/channel RAW scan, I could leave
some empty room on each side to allow for maximum flexibility for
myself and others in the future. Would this likely screw up the
initial exposure too far for later 16bpc adjustments to compensate?
It's most relevant for some very faded photos I'm scanning, in which
every slight contrast difference is relevant.

I'd guess good judgment of the maximum white point develops with
observation and experience. Clipping is bad indeed. I noticed that
a well-tuned histogram can spike, especially in the red channel, if
"saturation" is increased...annoying when more saturation seems
necessary. Another step to save until later? Or perhaps more user
error in the histogram.

BTW, thanks for your comments regarding Vuescan and Silverfast.
They're helping me revise how I consider the two products. One
thought I've had is that Vuescan's RAW scans plus a 16bpc Photoshop
equivalent can produce results fully equal or superior to
Silverfast's, along with the added flexibility of further 16bpc
editing. A fair comment? The only thing missing would be the
superior Digital ICE processing. It would be nice be able to apply
ICE to a 64bit RAW file or its separate channel equivalents; do you
know if Kodak supports this?

Please feel no obligation to respond to my long post. Thanks in
advance for any answers you do provide.

false_dmitrii

 

[false_dmitrii]

[Google's post-posting page didn't come up, so I'll resubmit this
once. Apologies if it shows up twice.]

1) can't comment on Epson in particular. Setting the curves individually
sounds like they are trying to remove any color casts. Limiting the

That seems to be the effect. The problem is that the varying
graypoint values make it more difficult and time-consuming to tinker
with the auto-exposure values. And there's no way to set
auto-exposure options.

output levels makes sense to me if the image doesn't have the full
dynamic range available and the software is trying to avoid posterizing
or they are trying to capture as much information as possible without
distorting what it perceives as neutral colors. The whitest white in the
image may not be 250,250,250 to the scanner.

Hmm, that makes sense. I checked again and found that the "200" upper
limit is for negatives, whereas prints limit white output to "245",
which was the value I originally associated with auto-exposure. Since
the values don't seem to vary, maybe they're hard-coded for particular
media types.

I had the impression that it was generally a good policy to capture as
much image data as possible when scanning images, including the
fullest possible contrast range, unless the goal is to minimize or
eliminate postscanning adjustments. A 16-bit/channel application can
then apply the output limits if necessary, no?

Is it unusual to scan negatives, with their inherent low contrast
range, using 245 or more output levels? For example, in a photo of a
tortoise, there was a large flash reflection on the tortoise's shell.
The 1-hour photo shop's print left a large "blowout" in this region,
and Epson Scan's auto-exposure did the same. Widening the output
lightness range and setting the color channels manually revealed
considerably more shell within the blowout zone. Similarly, Epson's
settings frequently produce a bright, almost-white sky in images where
more subtlety is present. Am I risking degradation of the image
colors by expanding the negatives' output levels past 200?

To put it another way, it seems to me that using close to the full
lightness range offers the best chance to preserve more of the full
range of the image's shadows and highlights as they would be perceived
in the real world, whereas limiting output levels to 200 means that
more of the perceived shadow and highlight detail must be sacrificed.

2) Usually, you can change the exposure to fill the histogram levels
pretty close to 0-255. A lot of people restrict the range to somewhere
around 245 to 250 levels. Some images won't have data values at the
extremes . Look out for clipping, where , when you look at the
histogram, there is a "pile" of pixels at either end. Changing the
exposure will help if your scanner hardware is capable.

My thinking was that with a 16-bit/channel RAW scan, I could leave
some empty room on each side to allow for maximum flexibility for
myself and others in the future. Would this likely screw up the
initial exposure too far for later 16bpc adjustments to compensate?
It's most relevant for some very faded photos I'm scanning, in which
every slight contrast difference is relevant.

I'd guess good judgment of the maximum white point develops with
observation and experience. Clipping is bad indeed. I noticed that
a well-tuned histogram can spike, especially in the red channel, if
"saturation" is increased...annoying when more saturation seems
necessary. Another step to save until later? Or perhaps more user
error in the histogram.

BTW, thanks for your comments regarding Vuescan and Silverfast.
They're helping me revise how I consider the two products. One
thought I've had is that Vuescan's RAW scans plus a 16bpc Photoshop
equivalent can produce results fully equal or superior to
Silverfast's, along with the added flexibility of further 16bpc
editing. A fair comment? The only thing missing would be the
superior Digital ICE processing. It would be nice be able to apply
ICE to a 64bit RAW file or its separate channel equivalents; do you
know if Kodak supports this?

Please feel no obligation to respond to my long post. Thanks in
advance for any answers you do provide.

false_dmitrii

 

[degrub]

That seems to be the effect. The problem is that the varying
graypoint values make it more difficult and time-consuming to tinker
with the auto-exposure values. And there's no way to set
auto-exposure options.




Hmm, that makes sense. I checked again and found that the "200" upper
limit is for negatives, whereas prints limit white output to "245",
which was the value I originally associated with auto-exposure. Since
the values don't seem to vary, maybe they're hard-coded for particular
media types.

I had the impression that it was generally a good policy to capture as
much image data as possible when scanning images, including the
fullest possible contrast range, unless the goal is to minimize or
eliminate postscanning adjustments. A 16-bit/channel application can
then apply the output limits if necessary, no?

Is it unusual to scan negatives, with their inherent low contrast
range, using 245 or more output levels? For example, in a photo of a
tortoise, there was a large flash reflection on the tortoise's shell.
The 1-hour photo shop's print left a large "blowout" in this region,
and Epson Scan's auto-exposure did the same. Widening the output
lightness range and setting the color channels manually revealed
considerably more shell within the blowout zone. Similarly, Epson's
settings frequently produce a bright, almost-white sky in images where
more subtlety is present. Am I risking degradation of the image
colors by expanding the negatives' output levels past 200?

Sounds like Epson is doing what the photolab does - satisfy most people
most of the time. There is no reason you cannot stretch the number of
levels, you just won't be able to use Epson's autoexposure. However,
with negatives, the lightest areas of the image (least transparent) will
be pushing the capabilities of the scanner to see real data. i think you
will just have to do it image by image and see how they look and develop
your own approach to get the maximum tonality from the scan.

To put it another way, it seems to me that using close to the full
lightness range offers the best chance to preserve more of the full
range of the image's shadows and highlights as they would be perceived
in the real world, whereas limiting output levels to 200 means that
more of the perceived shadow and highlight detail must be sacrificed.




My thinking was that with a 16-bit/channel RAW scan, I could leave
some empty room on each side to allow for maximum flexibility for
myself and others in the future. Would this likely screw up the
initial exposure too far for later 16bpc adjustments to compensate?
It's most relevant for some very faded photos I'm scanning, in which
every slight contrast difference is relevant.

I'd guess good judgment of the maximum white point develops with
observation and experience. Clipping is bad indeed. I noticed that
a well-tuned histogram can spike, especially in the red channel, if
"saturation" is increased...annoying when more saturation seems
necessary. Another step to save until later? Or perhaps more user
error in the histogram.

BTW, thanks for your comments regarding Vuescan and Silverfast.
They're helping me revise how I consider the two products. One
thought I've had is that Vuescan's RAW scans plus a 16bpc Photoshop
equivalent can produce results fully equal or superior to
Silverfast's, along with the added flexibility of further 16bpc
editing. A fair comment? The only thing missing would be the
superior Digital ICE processing. It would be nice be able to apply
ICE to a 64bit RAW file or its separate channel equivalents; do you
know if Kodak supports this?

ICE is only supported by a few flatbed scanners. i assume from your
questions that you have one of the recent Epson models that does, the
4870 ? Otherwise it will not be supported. i am not sure that the
flatbed versions are using infrared light. It may be more of a
diffraction or interference based algorithm. Vuescan and SF's algorithms
are pretty good but not perfect. They are based on software guesses with
user tuning. i suspect they do not soften the overall image as much as
ICE does and they certainly work with all film types.

Please feel no obligation to respond to my long post. Thanks in
advance for any answers you do provide.

My workflow (slides, prints) consists of setting white and black points
in the prescan (either auto or manual with SF or Vuescan) and scanning
16bpp at native resolution. i then embed an the IT8 scanner profile and
archive the image. After that i work in Photoshop or HDR if i used
Silverfast's HDR mode. Unfortunately, for negatives there is not an IT8
std and you either have to use Ed's film characterization (or
Silverfasts) or develop your own set of curves. THere is a method for
scanning negatives as postives and then inverting in PS or similar, but
i have not tried it.

A couple very good websites for reference and tutorials:

www.scantips.com

www.computer-darkroom.com

www.luminous-landscape.com/index.shtml

Also, i can heartily recommend Katrin Eismanns book
Photoshop Restoration and retouching

Frank

 

[SJS]

BTW, thanks for your comments regarding Vuescan and Silverfast.
They're helping me revise how I consider the two products. One
thought I've had is that Vuescan's RAW scans plus a 16bpc Photoshop
equivalent can produce results fully equal or superior to
Silverfast's, along with the added flexibility of further 16bpc
editing. A fair comment?

I am comparing Vuescan and Silverfast SE on my FS4000. With slides I
find that Vuescan produces a better result. My slides are mainly
Kodachrome and although the raw files from Vuescan seem too dark, the
detail is there and can be brought out with Photoshop. The Silverfast
scans are brighter but the dark detail is lost.

With negatives (mainly Fuji 400) I find the Vuescan scans are pale with
a green mask and lots of noise. With Silverfast the scans are much
brighter and cleaner. Silverfast is the definite winner here.

I'm not sure why I can't get Vuescan to match Silverfast with negatives.
This is a pity because I think Vuescan is better in many ways.

With my scanner it seems to be a case of horses for courses and quality
lost by using the wrong program for the film type is irrecoverable.

Regards,

Steven

 

[false_dmitrii]

false_dmitrii wrote:


ICE is only supported by a few flatbed scanners. i assume from your
questions that you have one of the recent Epson models that does, the
4870 ? Otherwise it will not be supported. i am not sure that the
flatbed versions are using infrared light. It may be more of a
diffraction or interference based algorithm. Vuescan and SF's algorithms
are pretty good but not perfect. They are based on software guesses with
user tuning. i suspect they do not soften the overall image as much as
ICE does and they certainly work with all film types.

Yes, I have a 4870. Whatever film scanner I end up buying will have
ICE, too. What I'm wondering is whether Kodak's standalone ICE
modules might be able to interact with the Vuescan or Silverfast RAW
infrared data. But I haven't checked with them directly yet.

Now that Silverfast has added ICE support, have you been able to use
ICE on a RAW file in HDR?

My workflow (slides, prints) consists of setting white and black points
in the prescan (either auto or manual with SF or Vuescan) and scanning
16bpp at native resolution. i then embed an the IT8 scanner profile and
archive the image. After that i work in Photoshop or HDR if i used
Silverfast's HDR mode. Unfortunately, for negatives there is not an IT8
std and you either have to use Ed's film characterization (or
Silverfasts) or develop your own set of curves. THere is a method for
scanning negatives as postives and then inverting in PS or similar, but
i have not tried it.

That sounds close to my desired approach. Nice that you can pick and
choose between the software. Epson's supplied all-purpose ICC
profile seems to work well enough for me; do you know whether Vuescan
Pro can use it?

A couple very good websites for reference and tutorials:

www.scantips.com

I've read this already. Quite helpful.

www.computer-darkroom.com

Focused on Photoshop and Silverfast, but informative about both.

www.luminous-landscape.com/index.shtml

I'll have to check this out.

Also, i can heartily recommend Katrin Eismanns book
Photoshop Restoration and retouching

I bought this after an earlier c.p.s. recommendation by "Hecate".
Lots of good advice and well-written to boot. It's helped fill in
some of the gaps left by my piecemeal Paint Shop Pro learning
experience (for anyone interested, most of the book's lessons are
easily applied to PSP, save for advanced color management, healing
brush, 16-bit adjustments, and other higher-end features). Though as
a non-painter, I was discouraged to see how often drawing skill comes
into play.

Thanks, Frank.
false_dmitrii

 

[Hecate]

I bought this after an earlier c.p.s. recommendation by "Hecate".
Lots of good advice and well-written to boot. It's helped fill in
some of the gaps left by my piecemeal Paint Shop Pro learning
experience (for anyone interested, most of the book's lessons are
easily applied to PSP, save for advanced color management, healing
brush, 16-bit adjustments, and other higher-end features). Though as
a non-painter, I was discouraged to see how often drawing skill comes
into play.

I'm glad you've found it useful. I had the first version but it got so
dog-eared I bought the latest version. I refer to it probably more
than any other book. I suspect I could repeat some of the advice in
the book verbatim, without looking

 

[degrub]

snip

Now that Silverfast has added ICE support, have you been able to use
ICE on a RAW file in HDR?

snip

ICE is only available in Ai, HDR supports their software SRD dust remover.


snip

That sounds close to my desired approach. Nice that you can pick and
choose between the software. Epson's supplied all-purpose ICC
profile seems to work well enough for me; do you know whether Vuescan
Pro can use it?

i believe that Vuescan needs to make it's own profile but i am not
certain. i have only used it that way. Google this newsgroup, there
were some threads on it when Ed first introduced IT8 calibration to Vuescan.

 

[degrub]

false_dmitrii wrote:

snip

Yes, I have a 4870. Whatever film scanner I end up buying will have
ICE, too. What I'm wondering is whether Kodak's standalone ICE
modules might be able to interact with the Vuescan or Silverfast RAW
infrared data. But I haven't checked with them directly yet.

From what i have seen, the ICE software is tied into the scanner code
directly and is not "standalone" like ROC or GEM . The scanner
manufacturer liscences the code from Kodak and includes it in the
scanner driver. i believe that Silverfast had to do that for Ai after a
number of users contacted them about the Nikon scanners. This was before
SF had fully developed SRD. Ed has shown no public inclination to
include ICE in his software that i am aware of.

 

[Ed Hamrick]

With negatives (mainly Fuji 400) I find the Vuescan scans are pale with
a green mask and lots of noise.

If you use the "Advanced Workflow Suggestions" in the VueScan
User's Guide, you can solve this problem.

Regards,
Ed Hamrick

 

[false_dmitrii]

From what i have seen, the ICE software is tied into the scanner code
directly and is not "standalone" like ROC or GEM . The scanner

Whoops, you're quite right. I had thought they had a software-only
version too.

SF had fully developed SRD. Ed has shown no public inclination to
include ICE in his software that i am aware of.

I'd imagine it's not especially cheap or easy to obtain...particularly
since they make such a point of providing it on a per-scanner basis,
whereas Vuescan is quite generous with its hardware support.
Otherwise, he could offer it as a separate upgrade. Hopefully,
someday Kodak will provide a hardware-independent post-scan version
that can read and work from an infrared channel.

false_dmitrii