Adjustments in TWAIN vs PaintShop Pro?

[Mark]

I am scanning photos with an Epson 1240U. The Epson TWAIN
software offers various tools (exposure, black and white points,
curves, gamma, unsharp mask, etc.) which are also available in
PaintShop Pro.

Is there any benefit to applying these corrections at scan time,
or will I do just as well to scan without adjustments and make
corrections later?

I am speculating that the TWAIN adjustments are just software
adjustments applied after the scan completes, and that any
difference between doing this in TWAIN or in PSP would only
result from the relative quality of the feature implementations in
the two pieces of software. But perhaps I am wrong about this?
Does the TWAIN driver have control over the scanner hardware
that would provide better adjustment results at scan time?

Thanks
-Mark

 

[CSM1]

I am scanning photos with an Epson 1240U. The Epson TWAIN
software offers various tools (exposure, black and white points,
curves, gamma, unsharp mask, etc.) which are also available in
PaintShop Pro.

Is there any benefit to applying these corrections at scan time,
or will I do just as well to scan without adjustments and make
corrections later?

I am speculating that the TWAIN adjustments are just software
adjustments applied after the scan completes, and that any
difference between doing this in TWAIN or in PSP would only
result from the relative quality of the feature implementations in
the two pieces of software. But perhaps I am wrong about this?
Does the TWAIN driver have control over the scanner hardware
that would provide better adjustment results at scan time?

Thanks
-Mark

The TWAIN interface does control the scanner hardware.
Exposure is controlled by TWAIN and scanner hardware. TWAIN is the software
that controls the scanner and interfaces to other imaging software. TWAIN is
machine dependent. Each scanner requires software to command the hardware to
do the work.

UnSharp Mask is probably best done in a good editor such as Photoshop.

I find that getting color balance right in the TWAIN interface is better
than correction in Photoshop from a bad color balance.
Some color correction in Photoshop is easier if the color balance is close
to start with.

You do want to get the most out of the film or photo that you can, so set
the scanner to get the best range of densities that are available in the
film or photo. (Black point and White point).

The Black and White points are very important when you scan.

 

[Mark]

You do want to get the most out of the film or photo that you can, so set
the scanner to get the best range of densities that are available in the
film or photo. (Black point and White point).

The Black and White points are very important when you scan.

So setting the black and white points in the TWAIN software will
definitely (or almost always) give me a better result than doing it
later in PaintShop Pro?

Thanks
-Mark

 

[CSM1]

So setting the black and white points in the TWAIN software will
definitely (or almost always) give me a better result than doing it
later in PaintShop Pro?

Thanks
-Mark

Yes.
You can not put back what was never there. If the data is lost in the scan,
you can not restore it in an Editor.

 

[AS]

In all "normal" flatbeds output is in 8 bits per channel. It is OK, if you
have it as a ready picture for your final output. It is not so good, if you
try to do lots of adjustments to get the result you want. It is especially
valid in getting details out of dark areas, adjusting black and white
points, correcting a huge color cast. Unsharp Mask and similar fine tuning
is done better in the final stage in dedicated editing software. So, if you
read carefully until now, you have your answer ;-)

Even if a scanner claims 16 bit/channel, you can't use it because of high
its own noise level. Literrary, you just get the noise in more bits, but the
useful signal is not so much better. It is not valid, if it's a professional
scanner with price of many thousand money bills. Though, people having such
a scanner usually know all this and don't ask questions ;-)

 

[Don]

Yes.

No. ;o)

In theory it makes no difference. In practice, it's far better to do
it in postprocessing later i.e., in your case using PaintShop Pro.

The reason is because in scanning software you usually have to work
with a tiny preview window (at preview resolution) instead of a full
blown image in an image editor (at full resolution), which image you
can also magnify. The number of tools in scanning software is usually
very minimal (if any), while an image editing program will have a full
complement of tools enabling you to evaluate the image much more
accurately

You can not put back what was never there. If the data is lost in the scan,
you can not restore it in an Editor.

That is correct. However, by setting black and white points in
scanning software you are, in effect, removing data (everything below
black and above white point). And you can't get this data back.

Since, as explained above, you can't really inspect the image in
scanning software, you are risking removing data which is not visible
in the scanning software but is visible in a(ny) image processing
software.

Therefore, if you care for maximum quality, it's advisable to limit
the amount of processing in the scanning software to a minimum e.g.,
only to hardware based corrections which can't be done later (e.g.,
ICE, hardware based grain removal, etc). All else should be done
later. Of course, this assumes you use maximum resolution and bit
depth.

Don.

 

[AS]

That is correct. However, by setting black and white points in
scanning software you are, in effect, removing data (everything below
black and above white point). And you can't get this data back.

Since, as explained above, you can't really inspect the image in
scanning software, you are risking removing data which is not visible
in the scanning software but is visible in a(ny) image processing
software.

True, but if you don't do this very aggresivly and leave some reserve, it
should be fine.

 

[Mark]

In theory it makes no difference. In practice, it's far better to do
it in postprocessing later i.e., in your case using PaintShop Pro.

Ok, my _theory_ is that the scanner hardware will try to provide
the best possible focus, exposure, and tonal range that it can
provide in a single scan. What motive could there be to do
otherwise? So the TWAIN controls should only provide a
better result than post-processing if there are hardware features
implemented (like Digital ICE,) or hardware-based tradeoffs in
the scanning process that must be determined at scan time
(again, through the TWAIN settings.) Everything else would be
a software adjustment post-scan, and should provide similar
results whether done at "scan time" in TWAIN, or weeks
later in PaintShop Pro.

Does that make sense, or am I off-base here?

Thanks
-Mark

 

[Mark]

True, but if you don't do this very aggresivly and leave some reserve, it
should be fine.

But is there any benefit to doing it in TWAIN at scan time, instead of
doing it later in an image editing program?

Thanks
-Mark

 

[Don]

But is there any benefit to doing it in TWAIN at scan time, instead of
doing it later in an image editing program?

If you set a couple of % tolerance as "AS" indicated, not really.
However, there are exceptions.

For example, Nikon's scanner software does all editing with 16 bits of
precision. Photoshop's so-called "16-bit" mode, on the other hand, in
reality only uses 15 bits!

So, in that particular case, using Nikon's scanner software to set B&W
point (as well as gamma, etc) would appear to provide more accuracy.

However, there's another twist. If your scanner only provides 14-bits
(or less) of image data then the above point becomes moot.

This may sound as nitpicking but at this level of detail it's really
important to make the distinction. It illustrates the complexities of
scanning and that there are all sorts of considerations one has to
take into account i.e. the context. (This "moving target" aspect of
scanning used to drive me nuts.)

But don't be discouraged by all this. In the end things do fall into
place but it takes time. If you do care about this level of detail
it's very common to re-scan your film collection several times as new
"revelations" are discovered. I've lost count of how many false starts
I had...

Don.

 

[Don]

Ok, my _theory_ is that the scanner hardware will try to provide
the best possible focus, exposure, and tonal range that it can
provide in a single scan. What motive could there be to do
otherwise?

They wouldn't do it intentionally but there are complications (focus,
for example - see below).

So the TWAIN controls should only provide a
better result than post-processing if there are hardware features
implemented (like Digital ICE,) or hardware-based tradeoffs in
the scanning process that must be determined at scan time
(again, through the TWAIN settings.)

Actually, hardware based features are not really better done at scan
time but the only time you can do them is at scan time. For example,
unless you set ICE (just like focus) at scan time, you simply can't do
it later. (In theory, you could save the IR channel data and emulate
ICE later, but that's beyond the scope of this discussion and will
only confuse the issue.)

Everything else would be
a software adjustment post-scan, and should provide similar
results whether done at "scan time" in TWAIN, or weeks
later in PaintShop Pro.

Does that make sense, or am I off-base here?

Basically that's correct with the caveats I mentioned last time (more
control in postprocessing). However, tonal (dynamic) range can be
improved by, for example, scanning twice - once for highlights and
once for shadows - or simply tailoring the range to the part of the
image where the object of interest is. Similarly, autofocus (like
anything else "auto") doesn't always work or may not set focus on the
part of the image you're interested in.

For example, I'm currently wrestling with old mounted slides. These
are heavily warped, meaning there is a considerable difference in
focus between the center of the slide and the corners. Now, if the
object of interest is in the center, autofocus will work fine. But if
the object is elsewhere, or you prefer an "average" focus, then
sampling and setting the focus manually will produce better results.

Finally, scanner software could include all the bells and whistles of
a dedicated image editing program. However, scanner software does not
because the manufacturers are usually mostly concerned with getting
the image out and provide only a minimum set of "extras" needed to get
a reasonable image. For most people (who don't even have an image
editing program) that's good enough, but if you're after a high
quality scan (and are willing to put in the time) then it's better to
limit scanner software adjustments to hardware features and do all
else later. That's known as a "raw" scan and is a whole different
subject in its own right.

BTW, (please ignore if you know this already) TWAIN itself is very
limited in what it can do. It just basically talks to the scanner and
gets the data. Stuff like ICE is really done at scanner software level
which technically sits "on top" of TWAIN. That's why I make a
distinction between TWAIN and scanner software.

Don.

 

[DavidTT]

Don, it seems like we are back to discussing raw scan again <g>.
Everything you said so far is right on. I will only add that on some
scanners that support hardware exposure control (by either lengthening
exposure time, and/or varying the light source intensity), a raw scan
should take advantage of this feature as needed.

 

[Mark]

Actually, hardware based features are not really better done at scan
time but the only time you can do them is at scan time. For example,
unless you set ICE (just like focus) at scan time, you simply can't do
it later. (In theory, you could save the IR channel data and emulate
ICE later, but that's beyond the scope of this discussion and will
only confuse the issue.)

Understood. And Digital ICE may be the _best_ example of
a hardware-based feature that must be taken advantage of at
scan-time, or not used at all.

However, tonal (dynamic) range can be
improved by, for example, scanning twice - once for highlights and
once for shadows - or simply tailoring the range to the part of the
image where the object of interest is.

Can you direct me to a tutorial that describes the process
of multiple-pass scanning for highlights and shadows?
That idea is new to me. It sounds intriguing.

I notice that Vuescan can instruct the scanner to make multiple
passes. What is the benefit here?

Similarly, autofocus (like
anything else "auto") doesn't always work or may not set focus on the
part of the image you're interested in.

I assume that manual focus is a feature of higher-quality scanners?
I don't think my Epson 1240U offers it.

For most people (who don't even have an image
editing program) that's good enough, but if you're after a high
quality scan (and are willing to put in the time) then it's better to
limit scanner software adjustments to hardware features and do all
else later. That's known as a "raw" scan and is a whole different
subject in its own right.

My preference is to do nothing at scan-time that I cannot do post-scan
with an image editing program. I just want to get as much useful
information as possible during the scan. If that is what a "raw scan" is,
then that is what I am looking for. As I mentioned before, one of my
limitations is that I have only temporary access to these family photos
and negatives, and I will probably not be able to get access again in
the future.

BTW, (please ignore if you know this already) TWAIN itself is very
limited in what it can do. It just basically talks to the scanner and
gets the data. Stuff like ICE is really done at scanner software level
which technically sits "on top" of TWAIN. That's why I make a
distinction between TWAIN and scanner software.

Interesting point. So the Epson "Twain 5" software that I am using
in "manual mode" is using the rudimentary TWAIN capability,
and adding a collection of software capabilities that are not part of
the TWAIN spec?

Thanks again,
-Mark

 

[Don]

Don, it seems like we are back to discussing raw scan again <g>.

It's hard to beat the advantages of a raw scan! ;o)

Everything you said so far is right on. I will only add that on some
scanners that support hardware exposure control (by either lengthening
exposure time, and/or varying the light source intensity), a raw scan
should take advantage of this feature as needed.

Yes! That's a good point!

Don.

 

[Don]

Can you direct me to a tutorial that describes the process
of multiple-pass scanning for highlights and shadows?
That idea is new to me. It sounds intriguing.

The procedure is generally known as "(digital) contrast masking" so if
you google for that you'll get quite a few hits. I'm not aware of a
tutorial as such because I did it the hard way, but I'm sure there is
one somewhere. Maybe someone else can chime in...?

The basics are as follows (again, please ignore if you know this):
The dynamic range of a scanner (i.e. the ability to expose both bright
and dark areas equally well) is often insufficient. The result of this
is that if you expose for highlights the shadow areas are a
featureless dark mess. Conversely, if you then boost exposure to bring
out the detail in shadows, the bright areas are all "burnt out" and
become a featureless white.

The idea is, therefore, to scan twice. Once, tailoring exposure to get
the most out of shadows, and then scan again but tailoring exposure to
get the most out of highlights. After that the two scans are combined
resulting in an image with equally well exposed dark and bright areas.

Well, that's the theory... ;o) In practice the first problem is
alignment of the two scans. Depending on how picky you are, you may
want or need to do subpixel alignment (i.e. move one image a fraction
of a pixel).

Once the images are aligned, you combine them. There are 3 ways of
doing this I'm aware of: using Gaussian Blur (google for "contrast
masking") or playing with Blending Modes in Photoshop (or similar
features in other software). The trouble (for me) in both of these is
that parts of the image I don't want "leak in".

So I devised my own method of only combining the parts of the image
I'm interested in. The problem there is that due to different
exposures the two images are no longer in "color sync", as I call it.
This is demonstrated by a visible border between the two images and a
general color mismatch. So you need to "color adjust" the two images
for a seamless combine. Again, this is too complicated to explain, but
essentially I sample narrow histogram "bands" and adjust the images
based on that.

I notice that Vuescan can instruct the scanner to make multiple
passes. What is the benefit here?

That's what VueScan tries to do (so-called "Long exposure pass").
However, I'm known to dislike VueScan intensely - to put it mildly..
;o)

The problems with VueScan are many. Even forgetting the numerous bugs
the basic flaw is that the images not aligned before they are
combined. Furthermore, when I tested VueScan a while back the "Long
exposure pass" results were quite mediocre with no visible
improvement.

This is different from multipass scans where the idea is to scan the
image using the same exposure only several times and then average out
the result. In theory this should improve the dynamic range. The catch
is that not all scanners support single-pass multiscanning in hardware
(move to next line, scan several times, then move to next line again,
etc). In the absence of *single-pass* multiscanning some software
tries to *multi-pass* multiscan (scan whole image, go back, scan
again, etc.). The problem with multi-pass multiscaning is that the
scans are not aligned so you don't really improve the dynamic range of
a scanner but only end up with a "soft" (slightly blurred) image.

I assume that manual focus is a feature of higher-quality scanners?
I don't think my Epson 1240U offers it.

I was really referring to film scanners. Flatbeds usually don't offer
focus or exposure controls. Focus is normally fixed to the top of the
glass.

My preference is to do nothing at scan-time that I cannot do post-scan
with an image editing program. I just want to get as much useful
information as possible during the scan. If that is what a "raw scan" is,
then that is what I am looking for.

Yes, that's exactly what a raw scan is!

As I mentioned before, one of my
limitations is that I have only temporary access to these family photos
and negatives, and I will probably not be able to get access again in
the future.

Most of us are in the same boat. Even if you do have access, photos
fade with time so it's best to get the most while you can.

Almost all people who prefer to raw scan usually backup these images
in a lossless format (e.g. as TIFF to CDR or DVD) before doing
anything to them. Often times, this is referred to as a "digital
negative". You then work on a copy to edit the image (crop, adjust
contrast, color, etc). The final product can then even be converted to
JPG for distribution or viewing. If - at a later date - you decide to
have another go at editing the image, you still have your raw digital
negative to edit everything from scratch without having to re-scan.

Interesting point. So the Epson "Twain 5" software that I am using
in "manual mode" is using the rudimentary TWAIN capability,
and adding a collection of software capabilities that are not part of
the TWAIN spec?

I'm not familiar with Epson software (maybe others can clarify) but I
have a sneaky suspicion that's simply what they call their scanner
software.

http://www.twain.org/ will tell you more about TWAIN but it's quite
technical.

Don.

 

[DavidTT]

It's hard to beat the advantages of a raw scan! ;o)

Right, and a topic seldom addressed in books and online tutorials.
That's why I suggested drafting a raw scan faq for this group a while
back, but never got around to it.

 

[DavidTT]

Can you direct me to a tutorial that describes the process
of multiple-pass scanning for highlights and shadows?
That idea is new to me. It sounds intriguing.

The procedure is generally known as "(digital) contrast masking" so if
you google for that you'll get quite a few hits. I'm not aware of a
tutorial as such because I did it the hard way, but I'm sure there is
one somewhere. Maybe someone else can chime in...?
[snip]

The basics are as follows (again, please ignore if you know this):
The dynamic range of a scanner (i.e. the ability to expose both bright
and dark areas equally well) is often insufficient. The result of this
is that if you expose for highlights the shadow areas are a
featureless dark mess. Conversely, if you then boost exposure to bring
out the detail in shadows, the bright areas are all "burnt out" and
become a featureless white.

The idea is, therefore, to scan twice. Once, tailoring exposure to get
the most out of shadows, and then scan again but tailoring exposure to
get the most out of highlights. After that the two scans are combined
resulting in an image with equally well exposed dark and bright areas.

Or, shoot the same image multiple times on a tripod, each time with a
different exposure. Then select and scan two (or more) frames to combine
in PS.

Well, that's the theory... ;o) In practice the first problem is
alignment of the two scans. Depending on how picky you are, you may
want or need to do subpixel alignment (i.e. move one image a fraction
of a pixel).

For me, this is *the* most difficult part of the process. The tutorials
are mostly for two exposures captured on digital cameras. These captures
are easier to align in PS since there is no scanner misalignment
involved.

 

[Don]

Right, and a topic seldom addressed in books and online tutorials.
That's why I suggested drafting a raw scan faq for this group a while
back, but never got around to it.

The main problem (for me at least) is time. I'm sure there's a lot of
good info in these threads but it takes to time to consolidate them...
:-(

It is interesting, though, how "natural" the concept of raw scans
seems. A large percentage of people who get into scanning
instinctively want to scan raw even though they may not even know what
to call it.

Don.

 

[Don]

For me, this is *the* most difficult part of the process. The tutorials
are mostly for two exposures captured on digital cameras. These captures
are easier to align in PS since there is no scanner misalignment
involved.

Did you catch my message on how I go about sub-pixel alignment in
Photoshop (although other software can be used also)? If you haven't
I'll search my archives and dig it up.

Even though sub-pixel alignment causes a very slight blurring it
basically solved the alignment problem for me. I "move" the image
which contains less data. Currently, that's the shadows and a slight
blurring there is not even noticeable. The only thing is that the
whole process, including combining the images afterwards, is time
consuming...

Don.