Scan Quality Vs Digital

[Kennedy McEwen]

Applying a blur filter is bad for small details.

True, but when you are comparing the result to a digital camera output,
even a very good digital camera, the blur is insignificant because it is
attenuating details that the digital camera just doesn't resolve anyway.
Remember, your 4000ppi scan on a full 35mm frame produces about 25Mpix,
the digital camera that the OP is comparing this to has an equivalent of
about 4million full colour pixels. So you can afford to lose quite a
lot of the finest detail in your scan before it becomes comparable.

 

[Kennedy McEwen]

I think you missed the point completely.

Worse than that Jon, I assumed that your post was an update from the OP
that his scanned film images were merely "a library of grain". Sorry,
my mistake.

Note to self: read post attributions before replying!!! ;-)

 

[Roger]

Which magic setting is the 'no grain' option on the LS-5000? I have an
LS-4000, but I have yet to get a scan of Gold 100 without any visible grain.

This may be due to semantics, but that isn't quite what I said. I
said I can not see the grain in the film. It's about the point where
the pixels in the scanner are the same size as the grain.

The following is just a random test of Velvia 100F:
http://misc.hq.phicoh.net/tmp1/Image3.jpg

That is definitely film grain as the patterns are quite visible, but
it looks a tad better than what I see in ASA 400.

I'll have to dig out some of the old scans to see what I can find. I
thought I had some up on the web, but all I found were digital. There
are some from ASA 400 but that has grain that is readily apparent even
in an inexpensive scanner.

Roger Halstead (K8RI & ARRL life member)
(N833R, S# CD-2 Worlds oldest Debonair)
www.rogerhalstead.com

 

[Hecate]

While the digital realm is making marked improvements it still has a
considerable way to go to get close to matching the performance of film.

I am a devout grain groupee, sometimes a pixel pony, slowly being made into
a mule cross.

;-) Yes! I have yet to see any digital process which approaches the
feel of my favourite B&W film for example (Tri-X). People say,
"oh,but it's got grain" and I say "yep, that's why I use it" ;-)

--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[David J. Littleboy]

;-) Yes! I have yet to see any digital process which approaches the
feel of my favourite B&W film for example (Tri-X). People say,
"oh,but it's got grain" and I say "yep, that's why I use it" ;-)

Don't you get tired of all your surfaces having the same texture (the
texture of the grain) and none of them having their own texture?

David J. Littleboy
Tokyo, Japan

 

[David J. Littleboy]

Your scan has much more visible grain that David's, that's what's wrong.

OK, David captured his from the frame edge, where the optic will nor
reproduce the grain as clearly, so that may be some of the issue,

I don't see any differences across the frame in grain. I suspect that if one
did see such, it would not be due to the optics, but to film flatness
issues. (I don't scan a frame if I can't get it flat within +/- 10 focus
units across the whole frame.)

I do see worse grain than that occassionally. Also, really high-contrast
edges often have a random pixel here or there that is off the wall bright,
and those get aggravated something fierce by sharpening.

I wonder how much good an antialiasing filter at the sensor would be. From
4000 ppi scans, I currently downsample to 2200 ppi and get pretty decent
images. I wonder if a 4800 dpi scanner with a 1.5 pixel shift AA filter
would provide images that could be downsampled to 2400 ppi to produce images
that are better per pixel than what I get now. That would make an upgrade
worthwhile. The problem is that most scanner users have the idea in their
heads that there is significant information in the upper octave of their
scans (to the best I can tell, there isn't any), so it wouldn't sell. Sigh.
But it should fix grain aliasing once and for all.

I interpret this to mean that you have corrected a yellow orange cast
produced at sunset, thus increasing the contrast in the blue channel and
bringing up the noise. So your image has been substantially post
processed - David's looks pretty raw.

Yep. I do a rough color balance in Nikon scan, scan into Picture Window Pro,
set black and white points and futz with the contrast in PWP's cutesy curves
control which shows both input and output histograms, drop to 24 bits and
save as a best quality JPEG.

But if there had been a strong yellow cast in my image, I might have
corrected it somewhat, but I make a point of not introducing clipping in my
to-be-archived jpegs.

The over saturated colours, (which are significantly clipped on all
channels) suggest that you may have been a little excessive with the post
processing. Hence your image is not a realistic representation of the
grain that is present on Velvia.

That seems likely. In a related area, I haven't figured out how to scan
negative materials well yet. Nikon scan produces a histogram that's empty in
the lower 1/4 or so of the range, and when I set the black point, the grain
noise gets ugly. The various heroic techniques of inverting the curves
oneself strike me as excessive beyond all reasonableness. (I'm using a very
old version of Nikonscan, and am loath to upgrade since it works fine for
slides.) Reports have it that the 9000 with the latest version of Nikonscan
does better, but the 9000 is US$3000 if you buy it in Japan....

David J. Littleboy
Tokyo, Japan

 

[Don]

I told you that a while back, when you were arguing that you couldn't
get accurate image statistics from the preview.

You did, indeed! If I find the time, I'd like to create a Nikon Scan
cache file by hand and stick a full resolution image (3946 x 5782) in
there to see what Nikon Scan does with it! ;o)

Will it just blow up or will the preview window suddenly sprout scroll
bars, or...?

In spite of everything, I find that the core of Nikon Scan is very
well programmed. As I said before, when I fed it a modified TWAIN
extended to +16 AG, Nikon Scan did not blink an eyelid and adjusted AG
exposure sliders accordingly. Excellent!! If only Nikon Scan were not
so bad tempered when I want to turn things off... ;o)

I preview at a
resolution close to 800x500 pixels on a 1600x1200 pixel display, so the
statistics and histograms are very accurate.

The problem was I only have 1024x768 here. But it's all a bit academic
now because I use my own program these days.

Hmmm. This might explain some strange effects I was getting while
experimenting with which profiles applied to which configurations a
while back.

I tried all permutations and this is what I got:

NKLS50_K.icm - Kodachrome
NKLS50_N.icm - negative, color
NKLS50_MN.icm - monochrome, negative
NKLS50_MN.icm - ditto + grayscale
NKLS50_P.icm - positive
NKLS50_P.icm - ditto + grayscale

So, grayscale is independent. Negative does not reverse but does
something else (?) and the same goes for Kodachrome.

Mind you, all that is the information stored in the preview cache.
Nikon Scan presumably does other things when one actually scans (as
you note below). Apparently a lot is done in TWAIN too (negative
processing i.e. orange mask removal, KC, etc).

For example, I'm still not sure where the color space profiles (sRGB,
Adobe, and friends) fit into this picture? Any idea?

The cache image itself is 16-bit, scanner profile (i.e. no profile or
"raw") and always stored in portrait orientation (i.e. as it comes off
the scanner). The cache file on my W2K installation lives at:

D:\Documents and Settings\All Users.WINNT\Application Data\Nikon\Nikon
Scan\4.0\Cache\CACHE001.DAT

There are flags for other things but I only decoded a few bits:

00000008: 80 00 00 00 - offset to strings (more below)
0000000C: 60 01 00 00 - preview width (here 352)
00000010: 04 02 00 00 - preview height (here 516)
....
00000028: 6A 0F 00 00 - total image width (3946 for LS-50)
0000002C: 96 16 00 00 - total image height (5782 for LS-50)
00000030: 03 00 00 00 - number of strings
....
' all strings are rounded up to 4-byte (longword) boundary
00000080: 14 00 00 00 - string length (including trailing 0 pad bytes)
00000084: 01 00 00 00 - string index
00000088: 4E 69 6B 6F... string bytes, here:
"Nikon COOLSCAN V ED" (scanner name, I guess)
'
0000009C: 3C 00 00 00 - string length (including trailing 0 pad bytes)
000000A0: 02 00 00 00 - string index
000000A4: 44 3A 5C 50... string bytes, here:
"D:\Program Files\Common Files\Nikon\Profiles\NKLS50_K.icm"
'
000000E0: 3C 00 00 00 - string length (including trailing 0 pad bytes)
000000E4: 03 00 00 00 - string index
000000E8: 44 3A 5C 50... string bytes, here:
"D:\Program Files\Common Files\Nikon\Profiles\NKLS50_R.icm"
....
00000124: ...
start of image data: 16-bit, interleaved RGB, no profile i.e. raw.

When the NS requires a specific profile and can't find it,
it throws up an error message. Scanning Kodachrome with the *_K.icm
profile removed still enabled a preview to be obtained, but a full scan
threw up an error. I guess if the preview didn't find the specific
profile it was looking for it defaulted to the second choice, *_R.icm,
and hence produced a preview without registering an error.

I don't know much about all this so ICC profile inspector data don't
mean much either.

However, from what I can see the "R" profile seems to set the black
point. It may do other things, but that's the most obvious. Could very
well be "R" is actually the "profile engine" and some flags in the
header tell it which profile(s) to apply (e.g. sRGB, Adobe, etc)?

Another strong indication that this is indeed the case is that when I
inspect a cache after previewing with "scanner profile" there is only
*one* string (the generic scanner identifier) but no *.icm strings!

Don.

 

[Philip Homburg]

Yep. I do a rough color balance in Nikon scan, scan into Picture Window Pro,
set black and white points and futz with the contrast in PWP's cutesy curves
control which shows both input and output histograms, drop to 24 bits and
save as a best quality JPEG.

But if there had been a strong yellow cast in my image, I might have
corrected it somewhat, but I make a point of not introducing clipping in my
to-be-archived jpegs.

Just for reference, here is a scan of the same frame without any additional
processing beyond ICE and 4x multi-pass:
http://misc.hq.phicoh.net/tmp1/Imaga3a.jpg

(I don't really archive scans. If I need scan for a good print, I re-scan.
The alternative would be to save the raw (16-bit/ch) scan as TIFF or PNG.)

In a related area, I haven't figured out how to scan
negative materials well yet. Nikon scan produces a histogram that's empty in
the lower 1/4 or so of the range, and when I set the black point, the grain
noise gets ugly. The various heroic techniques of inverting the curves
oneself strike me as excessive beyond all reasonableness. (I'm using a very
old version of Nikonscan, and am loath to upgrade since it works fine for
slides.) Reports have it that the 9000 with the latest version of Nikonscan
does better, but the 9000 is US$3000 if you buy it in Japan....

I gave up on 'negative mode' in NikonScan: scanning as positive gives
intuitive control of analog gain. Inverting and color balance in Photoshop
is quite easy.

 

[Hecate]

Don't you get tired of all your surfaces having the same texture (the
texture of the grain) and none of them having their own texture?

Quite the opposite. I find that good use of grain gives images an
almost tactile feel that the usual bland, smooth, digital shots don't
have. That doesn't mean that I exclusively use film, or that I don't
take digital images and convert them to B&W. Just that, for B&W, I
generally prefer film. You haven't lived until you've seen a
landscape taken on T-Max 3200 pushed to 6400 =- grain like golfballs


--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[David J. Littleboy]

Quite the opposite. I find that good use of grain gives images an
almost tactile feel that the usual bland, smooth, digital shots don't
have.

I'm not comparing to digital, I'm comparing to MF (and other people's LF).

The noise that grain introduces is a bogus irrelevancy. It wasn't in the
original scene, and I don't want it in my images.

David J. Littleboy
Tokyo, Japan

 

[differentstroles]

I'm not comparing to digital, I'm comparing to MF (and other people's LF).

The noise that grain introduces is a bogus irrelevancy. It wasn't in the
original scene, and I don't want it in my images.

You are entitled to prefer images without grain, but don't knock others'
preference. As an example, grain plays an important role in documentary
photography by suggesting realism and urgency. Digital shooters can only
achieve this by adding grain in Photoshop.

 

[Don]

I'm not comparing to digital, I'm comparing to MF (and other people's LF).

The noise that grain introduces is a bogus irrelevancy. It wasn't in the
original scene, and I don't want it in my images.

You're not scanning the original scene, you're scanning a film's
impression of that original scene, i.e. grain.

A flippant answer would be:
Then go back and re-shoot using a medium that doesn't have grain.

A constructive answer would be:
Wanting to recreate the original scene is a given, however, what is
also a given is that all that's left of that original scene is an
image on the film i.e. grain.

That's what's being scanned, not the original scene.

Don.

 

[David J. Littleboy]

You're not scanning the original scene, you're scanning a film's
impression of that original scene, i.e. grain.

Hmm. That makes no sense at all: I'm not scanning for the sake of scanning,
I'm scanning for the sake of making prints. And just as I never made grainy
prints when I did my own B&W printing in a darkroom, I never make grainy
prints when I scan.

A flippant answer would be:
Then go back and re-shoot using a medium that doesn't have grain.

Even better would be to do what I actually do: shoot with a medium that
doesn't have grain in the first place.

David J. Littleboy
Tokyo, Japan

 

[Kennedy McEwen]

I never make grainy
prints when I scan.

Yes you do - you posted a link to just such an example earlier in the
thread! ;-)

 

[David J. Littleboy]

Yes you do - you posted a link to just such an example earlier in the
thread! ;-)

Uh, Kennedy, that's a 6MP crop from a 93MP 6x7 scan: the grain's smaller
than the dot gain on any inket at anything under 16x20.

David J. Littleboy
Tokyo, Japan

 

[Kennedy McEwen]

Uh, Kennedy, that's a 6MP crop from a 93MP 6x7 scan: the grain's smaller
than the dot gain on any inket at anything under 16x20.

It's still grain, David, like it or not. Never is a long time. ;-)

 

[David J. Littleboy]

It's still grain, David, like it or not.

Not on the print it's not: even with a 20x loupe. (Yes, (a) I own a 20x
loupe, and (b) I've looked at prints with it.)

By the way, I'm having second thoughts about even 6x7. I downloaded the
first sample from the dpreview 5D preview gallery and printed a 6MP crop
from the 12.7MP image at A4. It looks very very good.

David J. Littleboy
(e-mail address removed)
Tokyo, Japan

 

[Hecate]

I'm not comparing to digital, I'm comparing to MF (and other people's LF).

The noise that grain introduces is a bogus irrelevancy.

....in your opinion...

It wasn't in the
original scene, and I don't want it in my images.

and that is *your* choice.

--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[Hecate]

Hmm. That makes no sense at all: I'm not scanning for the sake of scanning,
I'm scanning for the sake of making prints. And just as I never made grainy
prints when I did my own B&W printing in a darkroom, I never make grainy
prints when I scan.


Even better would be to do what I actually do: shoot with a medium that
doesn't have grain in the first place.

You're very mistaken if you're shooting film and assuming you don't
have any grain - you quite obviously do - it's just that you elect not
to emphasise it. I can get grainy images on a Pentax 67 quite easily
if I want to. Yet again, it8's *your* choice. Please don't dismiss
other people's choices just because you *personally* don't like it.
In case you didn't know, there's no one *right* way to make an image,
or to display and/or print it. Saying otherwise is just sheer
arrogance.

--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[Kennedy McEwen]

Not on the print it's not: even with a 20x loupe.

What is this magical printing technology that you use which does not
introduce its own grain, even with a 20x loupe? ;-)