Scanning very dark Kodachrome w/ coolscan V

[Dan]

I can't find a way to tie this to the thread from a month ago, but I
was having no luck scanning a very dark Kodachrome slide from 1978,
either in Vuescan or in Nikon scan. I tried everything suggested and
everything I could think of without success. I could barely see the
image in Vuescan, even when I raised the Analog gain. In Nikon scan, I
could get a very acceptable preview using ROC, but then the final scan
would never match the preview and would either be unusably dark or
filled with blue splotches.

Yesterday, I downloaded the 4.02 Nikon Scan update and, while a scan
from the same slide isn't perfect, I'm happy with the amount of detail
I was able to pull from it. Plus, the scan perfectly matched the
preview.

Thanks to all who gave me suggestions. I hope this helps anyone who's
still having a problem with dark Kodachromes.

Using Nikon Scan v.4.02, I raised the analog gain only slightly, used
dde for shadows and to cut down the blown out highlights (sky through
the forest), used digital ROC set to 2 and regular ICE. I also tweaked
the curves upwards and pulled down the Chroma curve to counteract
ROC's sometimes splotchy saturation. (w/o ROC, the slide still comes
out dark).

Dan

 

[Mike Engles]

I can't find a way to tie this to the thread from a month ago, but I
was having no luck scanning a very dark Kodachrome slide from 1978,
either in Vuescan or in Nikon scan. I tried everything suggested and
everything I could think of without success. I could barely see the
image in Vuescan, even when I raised the Analog gain. In Nikon scan, I
could get a very acceptable preview using ROC, but then the final scan
would never match the preview and would either be unusably dark or
filled with blue splotches.

Yesterday, I downloaded the 4.02 Nikon Scan update and, while a scan
from the same slide isn't perfect, I'm happy with the amount of detail
I was able to pull from it. Plus, the scan perfectly matched the
preview.

Thanks to all who gave me suggestions. I hope this helps anyone who's
still having a problem with dark Kodachromes.

Using Nikon Scan v.4.02, I raised the analog gain only slightly, used
dde for shadows and to cut down the blown out highlights (sky through
the forest), used digital ROC set to 2 and regular ICE. I also tweaked
the curves upwards and pulled down the Chroma curve to counteract
ROC's sometimes splotchy saturation. (w/o ROC, the slide still comes
out dark).

Dan

Hello

I think that ICE with dark Kodachromes is not a good idea. I think that
it causes the splodginess. I find that there is no great gradation
between min ROC and max. Having done ROC use the luninosity curve to
pull up the midtones. Finally try 2 or 3 GEM to clean up the graininess.

Roc is now much better now that they have fixed the preview not matching
the scan problem.

Mike Engles

 

[Don]

I think that ICE with dark Kodachromes is not a good idea.

Interestingly, due to lower resolution of the LS-30 ICE worked there
on Kodachromes without any problems (in my experience). However, on
the LS-50 the same images are unacceptable with ICE on!

ICE on Kodachrome has problems with high contrast areas. Instead of a
clearly defined transition viewed at 100% (or higher) a definitive
irregularity creeps in. Also, effects similar to "solarization" start
appearing on the dark side of the transition, which makes such images
unacceptable.

I did find some Kodachromes where ICE does work: Overexposed ones!
Presumably, this is because all the silver has been washed out and
there's nothing to confuse ICE.

One thing which is sorely missing from ICE is user control! If we had
the ability to adjust the cut-off between blemishes and image content
in the IR channel as well as the amount of correction, many
Kodachromes could be saved.

Better still, if ICE were a standalone utility intended for use with
RGBI images, now *that* would be ideal!

Making any global adjustments in NikonScan during scanning is really
not a viable, serious option especially without the ability to view
the image at 100% but having to work on that tiny preview. That's one
of the reasons why I, for one, don't even bother with ROC, GEM, DEE
etc.

Don.

 

[Kennedy McEwen]

Making any global adjustments in NikonScan during scanning is really
not a viable, serious option especially without the ability to view
the image at 100% but having to work on that tiny preview.

The preview doesn't *have* to be tiny. Just drag the window to fill as
big and area as you want - the entire screen if necessary. Until you
buy screen and video card capable of displaying 5,500 pixels or so
wide, it won't be a 100% view of course, but on the 1600x1200 display I
use, it isn't too bad at all. Certainly good enough to view most of the
defects that ICE conceals.

That's one
of the reasons why I, for one, don't even bother with ROC, GEM, DEE
etc.

I can't see what difference the size of the preview image makes at all
to perceiving the effect of ROC or DEE and, even on practical sized
displays, the effect of GEM is pretty obvious on preview.

 

[Erik Krause]

Better still, if ICE were a standalone utility intended for use with
RGBI images, now *that* would be ideal!

You can have this - with some personal effort - if you use VueScan. It
lets you save the RGBI data to disk. You can modify the I channel to
your needs and then process the images with vuescans IR-Clean option.
Or you make your own IR-Clean with some photoshop actions...

 

[Don]

The preview doesn't *have* to be tiny. Just drag the window to fill as
big and area as you want - the entire screen if necessary. Until you
buy screen and video card capable of displaying 5,500 pixels or so
wide, it won't be a 100% view of course, but on the 1600x1200 display I
use, it isn't too bad at all. Certainly good enough to view most of the
defects that ICE conceals.

Of course, the first thing I did is increase the Preview window as
much as I could, but as you say it's not 100%. And - I find - that to
do anything meaningful I need at least 100%, often more.
Otherwise, working in the Preview windows is like trying to tie your
shoelaces with your feet! ;o)

When it comes to defects ICE *introduces* to Kodachromes (KC) they can
only be observed once you start approaching 100%. Indeed, running an
A/B comparison of two KC scans - one with, and one without ICE - are
quite revealing at 100% and the effects are certainly totally
invisible in the Preview window. Had I not done that, I would have
wasted lots of time scanning KCs blissfully ignorant of what ICE was
doing to them.

So, conversely, I don't use any of the editing features of NikonScan
and just try to get as wide a dynamic range as possible (a piece of
cake with the LS-50!) and then edit (a copy) to my heart's content in
Photoshop later where I can see what I'm doing.

I can't see what difference the size of the preview image makes at all
to perceiving the effect of ROC or DEE and, even on practical sized
displays, the effect of GEM is pretty obvious on preview.

I ran some tests and they were pretty revealing.

At first blush, DEE looked "pretty good" in the preview window. I then
also scanned without DEE and, in Photoshop, used a simple and crude
Curves adjustment (just up and to the left). Both images were scanned
at 14 bits, of course.

A/B comparison in Photoshop (even at less than 100%!) revealed that -
after all the huffing and puffing - the DEE image was pretty mediocre
(too much contrast and very noisy) even when compared to the crude
Curves adjustment (smooth transitions and no additional noise).

So - at least for me - any adjustment at the scanning stage (other
than exposure) is not only "dangerous" (because it adulterates the
image at this very early stage) but this makes it also conceptually
"wrong" - at least for me - as there is no "digital negative". I
prefer to scan a "digital negative" first and then do adjustments (on
a copy) later where I can see what I'm doing as well as try different
adjustments at 100% without having to rescan.

As far as GEM is concerned there's a far superior Photoshop GEM
plug-in (more and finer controls) available for download from Kodak (a
watermarked trial version) and even with that the best I could do is
make the image look like when viewed through opaque plastic - and
that's making adjustments at 100%! So, I didn't even bother trying it
in NikonScan in the tiny Preview window.

Again - it's only me - but give me sharp beautiful grain over fuzzy
mess of a "grain dissolver" any day... Well, not that I like grain
particularly (unless it's in the cereal form) but you know what I
mean! ;o)

Don.

 

[Don]

You can have this - with some personal effort - if you use VueScan. It
lets you save the RGBI data to disk. You can modify the I channel to
your needs and then process the images with vuescans IR-Clean option.
Or you make your own IR-Clean with some photoshop actions...

Actually, I did try that - just for fun... The Photoshop alternative,
that is, although I didn't use actions but just played around by hand.

However, as someone here once wrote (I think it was Kennedy), ICE is
more than just applying threshold to the IR channel and blurring the
relevant areas. Anyway, it can be done in Photoshop, but I think it
would be very time consuming and labor intensive.

Instead, at least for Kodachromes which ICE doesn't like, I prefer to
correct the blemishes manually in Photoshop. It's also very time
consuming and labor intensive but on the plus side, the resulting
image is very sharp, although ICE4 is pretty good in this respect.

Of course, once I'm done with Kodachromes, I'll happily turn ICE back
on because with ICE4 the softening is minimal - barely noticeable.
Indeed, it appears to act as a mild grain dissolver without
"fuzzification"! On the one roll of non-Kodachromes slides I did so
far, ICE4 was really magical. I scanned all pictures twice (once with
and once without ICE) and compared them at 100% in Photoshop. ICE4 not
only removed all dust and scratches but it even smoothened the grain
somewhat without loss of sharpness! Perfect!

Don.

 

[Kennedy McEwen]

On Sat, 26 Jun 2004 23:11:02 +0100, Kennedy McEwen

Of course, the first thing I did is increase the Preview window as
much as I could, but as you say it's not 100%. And - I find - that to
do anything meaningful I need at least 100%, often more.
Otherwise, working in the Preview windows is like trying to tie your
shoelaces with your feet! ;o)

Doesn't agree at all with my own experience, but a lot depends on how
well the monitor and video card reproduce individual pixels - probably
also why you need more than 100% to do the job correctly.

So, conversely, I don't use any of the editing features of NikonScan
and just try to get as wide a dynamic range as possible (a piece of
cake with the LS-50!) and then edit (a copy) to my heart's content in
Photoshop later where I can see what I'm doing.

You can, of course, zoom in to 100% or greater in NikonScan using
exactly the same controls as in Photoshop - indeed, you can do exactly
the same in a preview window by cropping a smaller section of the image
before returning to the full frame for scanning in a single mouse click.

A/B comparison in Photoshop (even at less than 100%!) revealed that -

.... you *could* see the effect on an image of less than 100% scale! ;-)

Again - it's only me - but give me sharp beautiful grain over fuzzy
mess of a "grain dissolver" any day.

Careful with that axe Eugene!

Firstly, the term "grain dissolver" refers to a specific feature of the
top range Minolta scanner which is an optical technique and certainly
does not result in the distortion effect that you refer to. This was
originally developed under the term "Scanhancer" (see
http://scanhancer.iddo.nl) but adopted by Minolta and built in directly.
Rather than digitally filter the grain after it has been aliased and
distorted by the scanning process, the "grain dissolver" reduces the
contrast of spatial frequencies above the Nyquist limit of the scanner,
which are mainly grain, thus reducing the source of aliasing before it
actually occurs.

Secondly, there is no such thing as "sharp grain" on colour film
emulsion. It is a gelatinous dye globule. The larger grain which is
resolved is certainly soft and fuzzy in any case, the unresolved grain
is simply aliased, being unresolved by the sampling density. So what
you consider to be "sharp beautiful grain" is as much a scanner artefact
as the effect of excessive GEM.

Thirdly, the Kodak plug-in is similar in its results to the Nikon
supplied version, but much slower. The same is true of NeatImage.

 

[Mike Engles]

Actually, I did try that - just for fun... The Photoshop alternative,
that is, although I didn't use actions but just played around by hand.

However, as someone here once wrote (I think it was Kennedy), ICE is
more than just applying threshold to the IR channel and blurring the
relevant areas. Anyway, it can be done in Photoshop, but I think it
would be very time consuming and labor intensive.

Instead, at least for Kodachromes which ICE doesn't like, I prefer to
correct the blemishes manually in Photoshop. It's also very time
consuming and labor intensive but on the plus side, the resulting
image is very sharp, although ICE4 is pretty good in this respect.

Of course, once I'm done with Kodachromes, I'll happily turn ICE back
on because with ICE4 the softening is minimal - barely noticeable.
Indeed, it appears to act as a mild grain dissolver without
"fuzzification"! On the one roll of non-Kodachromes slides I did so
far, ICE4 was really magical. I scanned all pictures twice (once with
and once without ICE) and compared them at 100% in Photoshop. ICE4 not
only removed all dust and scratches but it even smoothened the grain
somewhat without loss of sharpness! Perfect!

Don.

Hello

The way round it is to do two scans, one with and one without ICE.
Layer them in Photohop with ICE under and then use a brush with layer
mask to remove the blemish and expose the ICE'd layer below.

Mike Engles

 

[Don]

You can, of course, zoom in to 100% or greater in NikonScan using
exactly the same controls as in Photoshop - indeed, you can do exactly
the same in a preview window by cropping a smaller section of the image
before returning to the full frame for scanning in a single mouse click.

That's very cumbersome for a number reasons: Having to switch back and
fore, not being able to tell exactly the magnification but only guess
and, most importantly, it's impossible to scroll around the magnified
image. Instead I'd need to rescan several times magnifying different
parts of the image, which makes comparison difficult without further
complications (copying and pasting the preview window in an editor,
for example). So, even though it's theoretically possible, it's not
really a viable option.

But as I say, doing all that editing at such an early stage in the
scanning process just seems conceptually and instinctively "wrong" to
me, unless speed is more important than quality in which case it's
fine doing it all in one bit swoop.

Careful with that axe Eugene!

Firstly, the term "grain dissolver" refers to a specific feature of the
top range Minolta scanner which is an optical technique and certainly
does not result in the distortion effect that you refer to.

I used the term in the generic sense. Replace with "grain reduction",
I don't think anyone copyrighted that yet... ;o)

This was
originally developed under the term "Scanhancer" (see
http://scanhancer.iddo.nl) but adopted by Minolta and built in directly.
Rather than digitally filter the grain after it has been aliased and
distorted by the scanning process, the "grain dissolver" reduces the
contrast of spatial frequencies above the Nyquist limit of the scanner,
which are mainly grain, thus reducing the source of aliasing before it
actually occurs.

I haven't tried Minolta's grain dissolver, so I can't comment on it,
but all GEM does for me is simulates putting opaque plastic over the
monitor. Again, it's probably only me, but I just don't find the
effect acceptable.

Secondly, there is no such thing as "sharp grain" on colour film
emulsion. It is a gelatinous dye globule. The larger grain which is
resolved is certainly soft and fuzzy in any case, the unresolved grain
is simply aliased, being unresolved by the sampling density. So what
you consider to be "sharp beautiful grain" is as much a scanner artefact
as the effect of excessive GEM.

Yes, I know about "grain clouds" by now, but putting all that aside
and simply viewing and judging two images side by side, one without
GEM and one with. Whatever the cause of the effect and no matter what
terminology is used, the one without GEM does look "sharper". In other
words, the cure seems worst than the disease! Well, to me, anyway.

Applying unsharp mask to the GEMed image seems to reduce its
effectiveness and sort of defies its purpose. You end up with grain
again, only twice removed and uneven. Again, it's probably personal
preference but by comparison I find the original image much more
acceptable.

Thirdly, the Kodak plug-in is similar in its results to the Nikon
supplied version, but much slower.

It does give much more control, though.


One other thing I noticed, this time regarding DEE. I'm quite
convinced it single-pass multiscans even on my LS-50. Of course,
that's nothing new, the scanner has the capability it's Nikon that
disabled it in software - I doubt very much it's disabled in firmware.

The reason I think DEE multiscans is because it's really a contrast
masking procedure (two scans). And, indeed, while trying extreme
values I noticed the same discoloration effect between highlights and
shadows I observed when contrast masking on my LS-30 so I think the
software works with two images scanned at different exposures.

Don.

 

[Kennedy McEwen]

On Sun, 27 Jun 2004 09:24:22 +0100, Kennedy McEwen


I haven't tried Minolta's grain dissolver, so I can't comment on it,
but all GEM does for me is simulates putting opaque plastic over the
monitor. Again, it's probably only me, but I just don't find the
effect acceptable.


Yes, I know about "grain clouds" by now, but putting all that aside
and simply viewing and judging two images side by side, one without
GEM and one with. Whatever the cause of the effect and no matter what
terminology is used, the one without GEM does look "sharper". In other
words, the cure seems worst than the disease! Well, to me, anyway.

You appear to be confusing a symptom of the problem for a desirable
feature - nothing wrong with that, some people like oversaturated
colours too. All grain reduction techniques will inevitably result in
some image softening, including the Scanhancer and Grain Dissolver. This
is because the scanner has more resolution than the sampling density can
support - it has a significant MTF at and beyond the Nyquist frequency -
and that results in aliasing. In effect, the image is *too* sharp for
the sampling density that is being used.

At 4000ppi most, but not all, of the image content lies below the
Nyquist frequency, however most of the grain lies above it, even on
relatively fast film, and so most of what aliases is grain. The
Scanhancer reduces this by lowering the MTF of the scanner above its
Nyquist, thus reducing the amount of grain that is resolved and aliased.
In doing so it also slightly reduces the MTF below Nyquist, but a
sharpening filter after the scan restores much of this - and only
applies to the frequencies present in the scan, so does not reintroduce
aliasing. Software techniques, however, attempt grain reduction after
the fact, and inevitably impact on real image data in the process since,
once aliased, the corrupted information cannot be distinguished from
real information. Consequently, it is a hit and miss affair, requiring
different settings of the software for different film types and
scanners. That is why the non-dedicated GEM is more controllable than
the integrated GEM in the Nikon scanner - it has to be to cope with the
different performance scanners used.

One other thing I noticed, this time regarding DEE.

Can't comment on DEE since my scanner doesn't support it and I have only
played with it a couple of times on other scanners.

 

[Don]

The way round it is to do two scans, one with and one without ICE.
Layer them in Photohop with ICE under and then use a brush with layer
mask to remove the blemish and expose the ICE'd layer below.

The problems is that at 4000 dpi the misalignment between the two
images is considerable. It was possible to do it at 2700 dpi (LS-30)
where I used it extensively to contrast merge, but at 4000 it's just
too much.

Now, I could do sub-pixel alignment but that additional and cumbersome
step further reduces the time advantage, already adding to the time
used for the second (ICE) scan.

The main problem is that ICE just doesn't work on some Kodachromes.
While it does remove, for example, small black specks of dust from
large areas of sky, it has a problem with high contrast transition
areas and I would still have to do a fair amount by hand.

So, in the end, just doing the whole job manually ends up being the
lesser of two evils.

Don.

 

[Hecate]

The problems is that at 4000 dpi the misalignment between the two
images is considerable. It was possible to do it at 2700 dpi (LS-30)
where I used it extensively to contrast merge, but at 4000 it's just
too much.

Now, I could do sub-pixel alignment but that additional and cumbersome
step further reduces the time advantage, already adding to the time
used for the second (ICE) scan.

There is an easier way:

Open first image, then second image. Move second onto first image so
they are in one file. Invert image on top layer and set opacity to
50%. Where the two images are misaligned you'll see what looks like
an embossed effect. Use the move tool to move the top image (and the
arrow keys when you get close to nudge pixel by pixel) until the
"emboss" disappears. The two are now aligned.

 

[Mike Engles]

There is an easier way:

Open first image, then second image. Move second onto first image so
they are in one file. Invert image on top layer and set opacity to
50%. Where the two images are misaligned you'll see what looks like
an embossed effect. Use the move tool to move the top image (and the
arrow keys when you get close to nudge pixel by pixel) until the
"emboss" disappears. The two are now aligned.

--

Hecate
(e-mail address removed)
veni, vidi, reliqui

Hello

In my experience it is not always so easy.
Sometimes there is a minute difference in the size of the iamges, so
that they cannot be completely aligned. Personally I really do not mind
if it is out by 1 pixel in 5000,others object to the inaccuracy.

Mike Engles

 

[Don]

You appear to be confusing a symptom of the problem for a desirable
feature - nothing wrong with that, some people like oversaturated
colours too.

No, no, I'm quite aware of it. That's why I said that the cure
(desirable feature) is worst than the disease (symptom) - again, all
judged by my subjective preferences. And, like they say, there is no
accounting for taste...

In effect, the image is *too* sharp for
the sampling density that is being used.

Exactly! I was taken aback (and was quite disappointed, actually) when
I saw my first slide scan. But after looking at the slide (or a
negative) with a magnifying glass, the "grain" was there all along. I
just didn't worry about it because it was masked by "fuzzification" of
photographic paper or massive enlargement when projecting slides. So,
in relative terms, the scanner simply reproduced faithfully what was
there. Because of that fact as well as having been staring at the
grain over the last year, I seem to have built up a "tolerance" and
I'm not as allergic to is as I initially was.

Software techniques, however, attempt grain reduction after
the fact, and inevitably impact on real image data in the process since,
once aliased, the corrupted information cannot be distinguished from
real information. Consequently, it is a hit and miss affair, requiring
different settings of the software for different film types and
scanners. That is why the non-dedicated GEM is more controllable than
the integrated GEM in the Nikon scanner - it has to cope with the
different performance scanners used.

That's a very good point. However, more control doesn't hurt and I
think it should be available in NikonScan as well.

But I guess that enters the "marketroid" arena. Having a plug-in with
"more knobs" may entice some people to buy GEM twice, once within the
scanner and once as a plug-in...

Can't comment on DEE since my scanner doesn't support it and I have only
played with it a couple of times on other scanners.

In my humble opinion, you're not missing much. Let me put it this way:
I'd trade DEE (and throw in GEM and friends) for your ability to
single-pass multi-scan any day! ;o)

I just came across a dark KC slide which needs it badly! Even though,
nominally, the full dynamic range is scanned, once I brighten up the
image sufficiently in Photoshop, dark areas are *very* uneven with
that notorious sprinkling of multi-colored pixels. Boosting exposure,
clears it up, but it also burns out the highlights. Shades of the
LS-30 just when I thought I was done with contrast masking... :-(

Don.

 

[Don]

There is an easier way:

Open first image, then second image. Move second onto first image so
they are in one file. Invert image on top layer and set opacity to
50%. Where the two images are misaligned you'll see what looks like
an embossed effect. Use the move tool to move the top image (and the
arrow keys when you get close to nudge pixel by pixel) until the
"emboss" disappears. The two are now aligned.

I used a similar procedure only instead of opacity I would try
different blending modes, for example, subtract one image from the
other. But I like your opacity idea better!

However, there are two additional problems, one mine and the other
conceptual:

1. I only have Photoshop 6 here and it doesn't support layers for
16-bit images.

I found a workaround, but it's cumbersome.

2. Most misalignments are sub-pixel i.e. one image needs to be nudged
a half or maybe a third of a pixel.

Again, I found a way to do that but it's also very messy and time
consuming.

Don.

 

[Erik Krause]

The problems is that at 4000 dpi the misalignment between the two
images is considerable. It was possible to do it at 2700 dpi (LS-30)
where I used it extensively to contrast merge, but at 4000 it's just
too much.

With vuescan you can do both scans from file. No alignment problems
here...

 

[Kennedy McEwen]

There is an easier way:

Open first image, then second image. Move second onto first image so
they are in one file. Invert image on top layer and set opacity to
50%. Where the two images are misaligned you'll see what looks like
an embossed effect. Use the move tool to move the top image (and the
arrow keys when you get close to nudge pixel by pixel) until the
"emboss" disappears. The two are now aligned.

That doesn't do sub-pixel alignement (unless you scale the images up
first) and certainly doesn't cope with "run out" in the scan direction,
resulting in the two images being slightly different sizes in one axis.

 

[Kennedy McEwen]

Exactly! I was taken aback (and was quite disappointed, actually) when
I saw my first slide scan. But after looking at the slide (or a
negative) with a magnifying glass, the "grain" was there all along. I
just didn't worry about it because it was masked by "fuzzification" of
photographic paper or massive enlargement when projecting slides. So,
in relative terms, the scanner simply reproduced faithfully what was
there.

On the contrary, the scanner is *emphasising* what is there in terms of
grain - which is why the "grain dissolver" works - it prevents that
emphasis.

 

[Hecate]

Hello

In my experience it is not always so easy.

Sometimes there is a minute difference in the size of the iamges, so
that they cannot be completely aligned. Personally I really do not mind
if it is out by 1 pixel in 5000,others object to the inaccuracy.

True. I find it works most of the time.