Scan Comparison Site, Call for Sample Scans
Rafe B.
I've assembled a collection of film scan samples at
<http://www.terrapinphoto.com/jmdavis/>
The idea is to see just how much image detail
can be pulled from 0.25" by 0.25" of film, or
an equivalent area of a digicam image.
Forget MTF curves and line pairs per millimeter.
This is a simple gallery of images achievable
in practice with good gear and good technique.
There are samples here from several popular
drum scanners (Aztec, Howtek, ICG) and CCD
scanners (Nikon, Minolta, Imacon, Leaf) As well
as a couple of digicams (Canon 1Ds, 10D.)
I'd love to post many more. Good sharp scans only.
Any camera, any lens, any film, any scanner.
Cherry pick the images all you like.
There are only a few simple rules:
1. Image should represent 0.25" by 0.25" of
film at native scan resolution.
2. No unsharp masking
3. Include a small overview image of the frame
from which the scan was taken and indicate the
film format (eg., 35 mm, 6x6, etc.)
4. Compress as high quality JPG.
5. Post to a web location and indicate the
URL, or email JPGs directly to me at:
rafe DOT bustin AT verizon DOT net
Please, before posting, have a look at the images
and scanners already represented. If your scanner
is already represented, there's no need to post
anything (and I will most likely ignore your submission)
unless it clearly matches or beats other samples from
that same model.
Thanks. And enjoy.
rafe b.
http://www.terrapinphoto.com
scan comparisons
http://www.terrapinphoto.com/jmdavis/
Paul Repacholi
...
> The idea is to see just how much image detail can be pulled from
> 0.25" by 0.25" of film, or an equivalent area of a digicam image.
> 4. Compress as high quality JPG.
JPEG will destroy the details you want to have in your samples, and
also remove much of the noise. This is the oposite of what you
need. You shoud use TIFF with data compression or the like. But *NO*
image compression!!
--
Paul Repacholi 1 Crescent Rd.,
+61 (08) 9257-1001 Kalamunda.
West Australia 6076
comp.os.vms,- The Older, Grumpier Slashdot
Raw, Cooked or Well-done, it's all half baked.
EPIC, The Architecture of the future, always has been, always will be.
David Chien
images scanned for comparison of contrast, resolution, etc.
Rafe B.
<pr...@prep.synonet.com> wrote:
>Rafe B. <rafe....@verizon.net> writes:
>
>...
>
>> The idea is to see just how much image detail can be pulled from
>> 0.25" by 0.25" of film, or an equivalent area of a digicam image.
>
>> 4. Compress as high quality JPG.
>
>JPEG will destroy the details you want to have in your samples, and
>also remove much of the noise. This is the oposite of what you
>need. You shoud use TIFF with data compression or the like. But *NO*
>image compression!!
I beg to differ. There is no observable difference in this
case between the JPG and the TIFF. No details have
been lost. Furthermore, TIFFs can't be viewed by web
browers and the file sizes of these images would be huge.
rafe b.
http://www.terrapinphoto.com
Ti Moute
I'm was reading about the scanned film vs digital images and it became
clear to me that I was confused about something that hopefully you can
clarify.
Why is it that "For digital captures, the camera's largest native
image size is taken to be 1" high"?
Thanks
Ti
Rafe B.
>I'm was reading about the scanned film vs digital images and it became
>clear to me that I was confused about something that hopefully you can
>clarify.
>
>Why is it that "For digital captures, the camera's largest native
>image size is taken to be 1" high"?
It's somewhat arbitrary, unless you're using 35 mm
as the reference, in which case it makes a good
deal of sense.
For starters, consider that 1.0" x 1.5" is a close
approximation to the true size of 35 mm image.
The 10D produces an image that's 2048 by 3072
pixels. There's no "dpi" figure in this case, but if
you equate this image to a 35 mm frame then you
can think of the 10D image as a "scan" at 2048 dpi.
Or think of it this way -- again, taking the 35 mm
frame to be 1" x 1.5" -- my arbitrary sample
images are 0.25" by 0.25" which is exactly
1/24 of a 35 mm frame. Applying this same
scaling to the 10D image you get a snippet
of 512 x 512 pixels -- ie., 1/24 of the total
10D image area.
So the "condensed" version of this arbitrary
rule is: the short axis of the digicam image,
measured in pixels, can be thought of as
the equivalent "dpi" of the digicam capture,
so long as we understand that we're
normalizing to a 35 mm frame that's taken
to be 1" by 1.5"
You might argue that this standard isn't
appropriate for scanning backs and MF
digital backs, but I'll cross that bridge
when I get some sample images from
one of those.
Paul Repacholi
>>Rafe B. <rafe....@verizon.net> writes:
>>...
Then you are wasting your time. You will have a collection that
is limited by whatever the JPEG encoder someone uses is, and
have lost the original data.
Perhaps you should get a better browser, mine if fine with TIFFs.
Ti Moute
If the digital image size of 1 inch is arbitrary, then I don't see how
it can be used for comparisons.
If the digital image size were arbitrarily 4 inches or 1/4 inch then
the results would more than likely be different on your site.
Wouldn't it be better to scan film at the same 2048 by 3072 pixels
with the same color depth then compare the two images.
Or either scan the film at maximum resolution and import the digital
image at its maximum then convert both to 1000x1000 for a comparison.
David J. Littleboy
"Paul Repacholi" <pr...@prep.synonet.com> wrote:
>
> > I beg to differ. There is no observable difference in this case
> > between the JPG and the TIFF. No details have been lost.
> > Furthermore, TIFFs can't be viewed by web browers and the file sizes
> > of these images would be huge.
>
> Then you are wasting your time. You will have a collection that
> is limited by whatever the JPEG encoder someone uses is, and
> have lost the original data.
I've never been able to find any significant differences between original
tiffs and low compression jpegs. Your comments are at odds with objective
reality. At sensible compression settings, the vast majority of pixels are
identical. Sure, at higher compression levels things get ugly, but not at
low compression settings. Given the small amount of high frequency
information (other than grain noise) in scans, the enormous tiff file macho
in scanner users is quite misplaced.
David J. Littleboy
Tokyo, Japan
Rafe B.
I'm afraid you quite missed the point.
This arbitrary one inch rule is only used in
an attempt to compare digicam captures to
film scans. It does not apply to the film scans
at all.
I've explained the logic of it, so I'm not going
to go through it again.
rafe b.
http://www.terrapinphoto.com
Rafe B.
<pr...@prep.synonet.com> wrote:
>Rafe B. <rafe....@verizon.net> writes:
>
>> On Tue, 16 Dec 2003 23:05:54 +0800, Paul Repacholi
>> <pr...@prep.synonet.com> wrote:
>
>>>Rafe B. <rafe....@verizon.net> writes:
>
>>>...
>
>>>> The idea is to see just how much image detail can be pulled from
>>>> 0.25" by 0.25" of film, or an equivalent area of a digicam image.
>
>>>> 4. Compress as high quality JPG.
>
>>>JPEG will destroy the details you want to have in your samples, and
>>>also remove much of the noise. This is the oposite of what you
>>>need. You shoud use TIFF with data compression or the like. But *NO*
>>>image compression!!
>
>> I beg to differ. There is no observable difference in this case
>> between the JPG and the TIFF. No details have been lost.
>> Furthermore, TIFFs can't be viewed by web browers and the file sizes
>> of these images would be huge.
>
>Then you are wasting your time. You will have a collection that
>is limited by whatever the JPEG encoder someone uses is, and
>have lost the original data.
>
>Perhaps you should get a better browser, mine if fine with TIFFs.
Trust me, nothing's been lost with the JPG conversion.
Those first two "perfect" scans are also JPGs and
yet it's still abundantly obvious how much clearer
they are than any of the film scans.
rafe b.
http://www.terrapinphoto.com
Rafe B.
>If the digital image size of 1 inch is arbitrary, then I don't see how
>it can be used for comparisons.
>If the digital image size were arbitrarily 4 inches or 1/4 inch then
>the results would more than likely be different on your site.
Yes, of course it would. But given that 35 mm is
presently the standard toward which most digicams
aspire, there is special significance to the dimensions
of the 35 mm frame -- and for simplicity's sake, I'm
calling that 1" by 1.5".
>Wouldn't it be better to scan film at the same 2048 by 3072 pixels
>with the same color depth then compare the two images.
>Or either scan the film at maximum resolution and import the digital
>image at its maximum then convert both to 1000x1000 for a comparison.
How do you scan film at 2048 x 3072 pixels?
You could scan film at 2048 dpi, which would give you
an image of around 2048 x 3072 pixels. And that would
in fact give you a fairly decent 8x10" print, if that were
true optical resolution we're talking about.
But 2048 is generally recognized to be on the low side
for scanning a sharp image on fine grained film. Even
I'll agree that film holds something more than that.
You are correct about one thing: a proper comparison
needs to recognize that different dpi scans (of course)
produce differently sized images from the same area
of film.
This is explained (albeit briefly) on the site. To "deal"
with it, simply choose some arbitrary and reasonable
standard image size and resample as necessary to
that size.
Eg. if you want to properly compare a Howtek scan
at 4399 dpi to the Nikon at 4000 dpi, one way would
be to upsample the Nikon scan sample from 1000 to
1100 pixels on a side.
rafe b.
http://www.terrapinphoto.com
Ti
says...
If you are talking about the information on the website, I'm not in agreement
with it but since you answered my question with two threads, I will continue
this in your other response with some reasons why.
Ti
says...
> On 17 Dec 2003 11:13:47 -0800, tim...@mchsi.com (Ti Moute) wrote:
>
>
> >If the digital image size of 1 inch is arbitrary, then I don't see how
> >it can be used for comparisons.
> >If the digital image size were arbitrarily 4 inches or 1/4 inch then
> >the results would more than likely be different on your site.
>
> Yes, of course it would. But given that 35 mm is
> presently the standard toward which most digicams
> aspire, there is special significance to the dimensions
> of the 35 mm frame -- and for simplicity's sake, I'm
> calling that 1" by 1.5".
>
for the sake of comparison. If you compared to other size film then you would
end up with different dpi for the digital camera. But since you are using 35mm
cameras, you have selected chosen 1" and I understand your reasoning but it
still doesn't mean it is factually correct.
>
> >Wouldn't it be better to scan film at the same 2048 by 3072 pixels
> >with the same color depth then compare the two images.
> >Or either scan the film at maximum resolution and import the digital
> >image at its maximum then convert both to 1000x1000 for a comparison.
>
>
> How do you scan film at 2048 x 3072 pixels?
>
> You could scan film at 2048 dpi, which would give you
> an image of around 2048 x 3072 pixels. And that would
> in fact give you a fairly decent 8x10" print, if that were
> true optical resolution we're talking about.
>
> But 2048 is generally recognized to be on the low side
> for scanning a sharp image on fine grained film. Even
> I'll agree that film holds something more than that.
ok
>
> You are correct about one thing: a proper comparison
> needs to recognize that different dpi scans (of course)
> produce differently sized images from the same area
> of film.
>
> This is explained (albeit briefly) on the site. To "deal"
> with it, simply choose some arbitrary and reasonable
> standard image size and resample as necessary to
> that size.
>
> Eg. if you want to properly compare a Howtek scan
> at 4399 dpi to the Nikon at 4000 dpi, one way would
> be to upsample the Nikon scan sample from 1000 to
> 1100 pixels on a side.
Here is your first words on your page "How much image detail from 0.25" by 0.25"
of film?"
The would think best way to determine detail would be to see how big a picture
could be produced from the film without losing its quality. The same would go
for a digital image.
Going the other way would seem to determine the quality of the scanner.
Bob Monaghan
Unless you use the loss-less mode of JPEG (which doesn't give much
compression), you are losing data and can't recreate the original file or
image.
As David L. pointed out, this lost data is mostly high frequency data.
This is precisely the same data which provides high contrast and high
resolution aspects to images. It happens that high frequency data has a
LOT more information than the base image data, i.e., perhaps 75%+ of the
image data may be high frequency data in a film based image and absent in
a 11 MP direct capture digital image.
The difference is equiv. to putting a fog filter over a very good lens;
the high frequency data is blocked. The various 3MP or 6MP or 11 MP
systems are just different levels of diffusion or fog filters or low pass
filters.
How is that 75% of high frequency data being lost? One loss source is the
use of compression algorithms like JPEG which achieve compression by
reducing the high frequency components. So after you process a film scan
with compression algoriths, you have generally removed most or all of the
high frequency data in the original image (hence, you get compression
"quality" options, based on how much you want to compress and how much
data you are willing to permanently lose).
That's why folks want to see the original files, with larger being a
necessary evil in order to deal with all of the original data.
Suppose I processed an 11 MP direct capture image until it was as small as
a 2 MP image, and they looked the same. Would you then agree that the 11
MP image is no better than the 2 MP image? Probably not, huh? But that's
what you are doing with the film image, which is ~40 MP (100 ISO). If you
weren't doing so, then you wouldn't be able to say it was just as good or
not quite as good as an 11 MP digicam image ;-)
regards bobm
--
***********************************************************************
* Robert Monaghan POB 752182 Southern Methodist Univ. Dallas Tx 75275 *
********************Standard Disclaimers Apply*************************
Rafe B.
>Here is your first words on your page "How much image detail from 0.25" by 0.25"
>of film?"
>
>The would think best way to determine detail would be to see how big a picture
>could be produced from the film without losing its quality. The same would go
>for a digital image.
>
>Going the other way would seem to determine the quality of the scanner.
What is so difficult to comprehend about this?
In what way does my methodology compromise
or "constrain" image quality?
I've posted scans at full resolution based on
a fixed area of film.
It goes without saying (or so I thought) that larger
film formats produce more total information.
Choosing 1/4" by 1/4" was merely a convenient
reference area, for several reasons:
1. With the best film scanners typically
returning images of 4000 dpi, the resulting
snippet size is 1000 x 1000 pixels, which
can be viewed on a decent hi-res monitor.
2. Hopefully obvious, but using a small
area also allows users to scan at the full
capabilities of their scanners.
2. Bandwidth issues for web browsing
3. File size issues for web browsing
Rafe B.
wrote:
>
>Unless you use the loss-less mode of JPEG (which doesn't give much
>compression), you are losing data and can't recreate the original file or
>image.
>
>As David L. pointed out, this lost data is mostly high frequency data.
>This is precisely the same data which provides high contrast and high
>resolution aspects to images. It happens that high frequency data has a
>LOT more information than the base image data, i.e., perhaps 75%+ of the
>image data may be high frequency data in a film based image and absent in
>a 11 MP direct capture digital image.
bob, you really ought to get out of your ivory tower
every now and then.
Now, those two "perfect" scans are encoded in
JPG and yet contain OODLES more high-frequency
data than any of the actual film scans.
There is no significant loss of image data in
the film scan snippets posted. Your numbers
don't mean diddly here.
>The difference is equiv. to putting a fog filter over a very good lens;
>the high frequency data is blocked. The various 3MP or 6MP or 11 MP
>systems are just different levels of diffusion or fog filters or low pass
>filters.
>
>How is that 75% of high frequency data being lost? One loss source is the
>use of compression algorithms like JPEG which achieve compression by
>reducing the high frequency components.
No, that's absolutely not true. JPG compression is
due to two factors:
1. Decimation of color information, after
converting the image to a Lab or Yuv
color space.
2. Lowering the resolution (by user-selectable
factors) of the coefficients of the DCT of the image.
>So after you process a film scan
>with compression algoriths, you have generally removed most or all of the
>high frequency data in the original image (hence, you get compression
>"quality" options, based on how much you want to compress and how much
>data you are willing to permanently lose).
>
>That's why folks want to see the original files, with larger being a
>necessary evil in order to deal with all of the original data.
>
>Suppose I processed an 11 MP direct capture image until it was as small as
>a 2 MP image, and they looked the same. Would you then agree that the 11
>MP image is no better than the 2 MP image? Probably not, huh? But that's
>what you are doing with the film image, which is ~40 MP (100 ISO). If you
>weren't doing so, then you wouldn't be able to say it was just as good or
>not quite as good as an 11 MP digicam image ;-)
Forget the freaking digicams. Forget the 40 megapixels
too, while you're at it. Red herrings.
Viewed on a monitor or printed, if I apply high quality JPG
encoding to one of my 20 Mpixel 35 mm film scans, or
one of my 60 Mpixel 645 scans, there is absolutely no
visible difference on print or on screen.
I've posted a TIF corresponding ot one of the LS-8000
JPG scan snippets.
Here's the TIF:
<http://www.terrapinphoto.com/jmdavis/chrome_41_tif.tif>
and here's the JPG:
<http://www.terrapinphoto.com/jmdavis/chrome_41_ice.jpg>
In terms of viewable detail, there's not one iota of
difference between these two images.
Ya know, I sure wish you'd send me a gorgeous scan
sample or two instead of trying to poke holes in my
methodology with irrelevant and misapplied theory.
rafe b.
http://www.terrapinphoto.com
scan samples
http://www.terrapinphoto.com/jmdavis
David J. Littleboy
"Bob Monaghan" <rmon...@engr.smu.edu> wrote:
>
> Unless you use the loss-less mode of JPEG (which doesn't give much
> compression), you are losing data and can't recreate the original file or
> image.
JPEG (prior to JPEG 2000) does not have a lossless mode.
> As David L. pointed out, this lost data is mostly high frequency data.
So far so good, but you have misunderstood (a) the magnitude of the loss
(insignificant) and (b) the amount of high-frequency _information_ present
in film scans (also insignifcant).
> This is precisely the same data which provides high contrast and high
> resolution aspects to images. It happens that high frequency data has a
> LOT more information than the base image data, i.e., perhaps 75%+ of the
> image data may be high frequency data in a film based image and absent in
> a 11 MP direct capture digital image.
Film images have almost no high frequency information in them at all.
Between lens MTF (low at 40 lp/mm over the majority of the frame) and film
MTF (Provia is 50% MTF at 35 lp/mm), there simply isn't significant high
frequency information. Grain noise swamps high frequency information for all
but the highest contrast targets at anything over 40 lp/mm. (At 4000 dpi and
over grain noise means that scans have less than 6 bits of valid data.)
It's widely understood that the significant image content in film images is
in the 20 lp/mm range.
> The difference is equiv. to putting a fog filter over a very good lens;
> the high frequency data is blocked. The various 3MP or 6MP or 11 MP
> systems are just different levels of diffusion or fog filters or low pass
> filter.
You've obviously never compared a tiff out of a scanner or digital camera to
a low compression jpeg of the same image.
_There are no visually detectable differences_
Try it. Create a tiff from your favorite scanner and sharpest slide, and
save it as a jpeg at various compression levels. Print them out at 300 dpi,
and hand them to people and ask them to tell which is which. Subtract the
numerical values in Photoshop and tell us what changed.
> How is that 75% of high frequency data being lost?
It's not being lost. There is no visually detectable difference between
tiffs and low-compression jpegs. The differences are far smaller than the
level of the noise in the original image.
stacey
> Furthermore, TIFFs can't be viewed by web
> browers
??? Even the basic KDE browser Konquerer opens them fine. Maybe you need a
better browser/OS?
--
Stacey
stacey
>
> I've posted a TIF corresponding ot one of the LS-8000
> JPG scan snippets.
>
> Here's the TIF:
> <http://www.terrapinphoto.com/jmdavis/chrome_41_tif.tif>
> and here's the JPG:
> <http://www.terrapinphoto.com/jmdavis/chrome_41_ice.jpg>
>
>
> In terms of viewable detail, there's not one iota of
> difference between these two images.
>
??? You must be blind if you think these look the same! How can you use info
like this as a display of what different scans look like? The tiff looks
like a pretty nice scan, the jpg looks like a noisy mess that would make me
think it's unusable. Maybe that's your point, to -make- film look worse
than your loved digicams? Thanx for showing this so I know how bogus this
info you're posting is..
--
Stacey
Rafe B.
No, thanks for showing us your bias, ignorance, and
failure to remain on topic.
The JPG's got a bit more contrast is all, which tends to
accentuate the grain. They're both from the same exact
scan. bobm's argument has to do with high frequency
detail. Not at all about tonality, which is the only
difference between these two images.
Get it through your brain, if you can, that this is from
0.0625 square inches of film. Or show us that you can
do better, from the same area of film.
And this particular discussion has *nothing* to do with
digicams, so why bring it up?
rafe b.
http://www.terrapinphoto.com
stacey
> On Thu, 18 Dec 2003 22:36:27 -0500, stacey <foto...@yahoo.com> wrote:
>
>>Rafe B. wrote:
>>
>>>
>>> I've posted a TIF corresponding ot one of the LS-8000
>>> JPG scan snippets.
>>>
>>> Here's the TIF:
>>> <http://www.terrapinphoto.com/jmdavis/chrome_41_tif.tif>
>>> and here's the JPG:
>>> <http://www.terrapinphoto.com/jmdavis/chrome_41_ice.jpg>
>>>
>>>
>>> In terms of viewable detail, there's not one iota of
>>> difference between these two images.
>>>
>>
>>
>>??? You must be blind if you think these look the same! How can you use
>>info like this as a display of what different scans look like? The tiff
>>looks like a pretty nice scan, the jpg looks like a noisy mess that would
>>make me think it's unusable. Maybe that's your point, to -make- film look
>>worse than your loved digicams? Thanx for showing this so I know how bogus
>>this info you're posting is..
>
>
> No, thanks for showing us your bias, ignorance, and
> failure to remain on topic.
>
> The JPG's got a bit more contrast is all,
A bit? Look at the brown wall if front of the buiulding In the tiff, it's
pretty smooth, in the jpg it's a noisy mess!
> which tends to
> accentuate the grain. They're both from the same exact
> scan.
But they -aren't- the same at all, you've changed what the original film
scan looks like for your display.
>
> And this particular discussion has *nothing* to do with
> digicams, so why bring it up?
>
Uh because you have digital capture compared to film? Somehow I'm sure you
will be doctoring the images to support your love for digicams? These film
scans you -ruined- are evidence of this. Again thanx for showing what
you're up to, shame people will think you are being up front about what is
really going on in your "display".
--
Stacey
Rafe B.
>> The JPG's got a bit more contrast is all,
>
>A bit? Look at the brown wall if front of the buiulding In the tiff, it's
>pretty smooth, in the jpg it's a noisy mess!
It's grain, Stacey. All film has it. This is from
a chrome, albeit a very old one.
>> which tends to
>> accentuate the grain. They're both from the same exact
>> scan.
>
>But they -aren't- the same at all, you've changed what the original film
>scan looks like for your display.
Your only point is tonality. Yes, the TIF was
pulled later -- in the context of bobm's complaints
about detail -- and with slightly less contrast
than the JPG. I've since fixed that; the JPG
and TIF are now from exactly the same snip.
Changes in contrast will affect appearance
of grain, and that's just as true in analog/wet
darkroom as in film scanning and Photoshop.
>> And this particular discussion has *nothing* to do with
>> digicams, so why bring it up?
>>
>
> Uh because you have digital capture compared to film? Somehow I'm sure you
>will be doctoring the images to support your love for digicams? These film
>scans you -ruined- are evidence of this. Again thanx for showing what
>you're up to, shame people will think you are being up front about what is
>really going on in your "display".
You're completely and utterly off base, stacey.
rafe b.
http://www.terrapinphoto.com
stacey
> On Thu, 18 Dec 2003 23:30:08 -0500, stacey <foto...@yahoo.com> wrote:
>
>
>>> The JPG's got a bit more contrast is all,
>>
>>A bit? Look at the brown wall if front of the buiulding In the tiff, it's
>>pretty smooth, in the jpg it's a noisy mess!
>
> It's grain, Stacey. All film has it. This is from
> a chrome, albeit a very old one.
I've looked at my chromes closely and it's funny, they don't have
multicolored "grain" like these scans you post do and claim this noise is
"grain".
>
> Your only point is tonality. Yes, the TIF was
> pulled later -- in the context of bobm's complaints
> about detail -- and with slightly less contrast
> than the JPG. I've since fixed that; the JPG
> and TIF are now from exactly the same snip.
And "changing" them makes for a equal comparison??
--
Stacey
Rafe B.
>Rafe B. wrote:
>
>> On Thu, 18 Dec 2003 23:30:08 -0500, stacey <foto...@yahoo.com> wrote:
>>
>>
>>>> The JPG's got a bit more contrast is all,
>>>
>>>A bit? Look at the brown wall if front of the buiulding In the tiff, it's
>>>pretty smooth, in the jpg it's a noisy mess!
>>
>> It's grain, Stacey. All film has it. This is from
>> a chrome, albeit a very old one.
>
>I've looked at my chromes closely and it's funny, they don't have
>multicolored "grain" like these scans you post do and claim this noise is
>"grain".
Show me a scan of a print from one of your
chromes representing the same image area:
0.0625 square inches of film. Then we'll talk.
>> Your only point is tonality. Yes, the TIF was
>> pulled later -- in the context of bobm's complaints
>> about detail -- and with slightly less contrast
>> than the JPG. I've since fixed that; the JPG
>> and TIF are now from exactly the same snip.
>
>And "changing" them makes for a equal comparison??
If we're talking about high frequency content or
spatial detail, it's absolutely insignificant.
The change wasn't deliberate, and wasn't
intended to obfuscate, and has since been
corrected. That's it. Good night.
rafe b.
http://www.terrapinphoto.com
Alan Browne
Rafe B. wrote:
> I beg to differ. There is no observable difference in this
> case between the JPG and the TIFF. No details have
> been lost. Furthermore, TIFFs can't be viewed by web
> browers and the file sizes of these images would be huge.
Netscape 7 accesses TIF's via a viewer such as Quicktime.
jjs
<alan....@FreeLunchVideotron.ca> wrote:
And TIFFs can be compressed.
Ti
says...
your examples. When I look at your examples, they do not equate to 1/4 of the
scanned film if the 35mm was 1" x 1" nor 1/6 of the film if it were 1" x 1.5"
examples of what I am talking about:
1)
On the Aztek Drum Scans -- "Standard" 0.25 x 0.25 snippets
Photos and scans by J. Arthur Davis
The Overview 4x5 chrome, I see the highlighted area to be much smaller than 1/4
or 1/6 of the entire image. If this is 1" x 1" of a 4" x 5" photo, then you are
not using 1/4 of the film but only 1/20 of the image. If you are only using 1/4"
x 1/4" of a 4" X 5" photo then you are only looking at 1/80 of the entire image.
If this is the case, then the digital image should also be taken at the same
ratios.
2)
On the Howtek 4500 Drum scan vs. LS-8000
Two Scans of the same image
Photo and scans by Lawrence Smith
The Overview 35 mm, Kodak 100sw pushed two stops, the highlighted areas are not
1/4 or 1/6 of the total image.
3)
On the Heidelberg Tango Drum vs LS-8000
Two Scans of the same image
Photo and scans by Bill Hilton
The Overview 645, Velvia, the boxed image is not 1/4 or 1/6 of the entire image.
Giving these 3 examples, can you see why I am confused when you say "How much
image detail from 0.25" by 0.25" of film?" which would be either 1/4 or 1/6 of
the entire image when none of the images are 1/4 or 1/6 of the image.
Bill Hilton
>I comprehend your methodology and perhaps I am not getting a clear picture
>on your examples.
You apparently are assuming all the films were 35 mm when in fact most of them
were not.
> When I look at your examples, they do not equate to
>1/4 of the scanned film if the 35mm was 1" x 1" nor 1/6 of the film if it were
>1" x 1.5"
>
>examples of what I am talking about:
>
>3)
>On the Heidelberg Tango Drum vs LS-8000
>Two Scans of the same image
>Photo and scans by Bill Hilton
>The Overview 645, Velvia, the boxed image is not 1/4 or 1/6 of the entire
>image.
645 refers to the film size being scanned, ie, 6 x 4.5 cm or about 2.7x larger
than 35 mm. It's "not 1/4 or 1/6 of the entire image" because the film is much
larger than 35 mm. .25x.25" would only be 1/4 the image if the image (film)
were 1x1".
>1)
>On the Aztek Drum Scans -- "Standard" 0.25 x 0.25 snippets
>Photos and scans by J. Arthur Davis
>The Overview 4x5 chrome, I see the highlighted area to be much smaller than
>1/4 or 1/6 of the entire image.
That's because the film is 4x5", not 1x1.5" (which is what 35 mm is, roughly).
>2)
>On the Howtek 4500 Drum scan vs. LS-8000
>Two Scans of the same image
>Photo and scans by Lawrence Smith
>The Overview 35 mm, Kodak 100sw pushed two stops, the highlighted areas are
>not 1/4 or 1/6 of the total image.
I think this one was 6x7 cm or more than 4x larger than 35 mm.
Rafe picked .25" because it would fit on the screen so long as the scan was
4000 dpi or less. This was a very sensible move, it seems to me.
Bill
Bill Hilton
>
>> Choosing 1/4" by 1/4" was merely a convenient
>> reference area, for several reasons:
>From: Ti T...@hotmail.com
>
> When I look at your examples, they do not equate to 1/4 of the
>scanned film if the 35mm was 1" x 1"
A .25x.25" snippet would equate to 1/16th not 1/4th of a 1x1" film or 1/24th of
35 mm film (~ 1 x 1.5").
Bill
Bob Monaghan
a) JPEG has a loss-less mode option, which provides relatively little
compression, but enables you to recreate the original file without
losses; and it has long been part of the full JPEG standard
b) it is correct that much of the color space is being compressed,
throwing away color data to save space. This is one way in which original
photos (and 48 bit scans) should be superior to digital prints in color
fidelity.
c) despite the claim of insignificant contrast above 35 lpmm on film,
reports of test results by lens testing services routinely report lens
high aerial contrast (as in 60 to 80%+) and high resolutions, centrally as
high as 450 to 600+ lpmm aerial resolution (E.g., pentax 50mm and leica
50mm respectively). System resolution calculations produce resolutions
over 100 lpmm, which is what we observe in actual tests. Zeiss tests at
1996 photokina exceeded 200 lpmm in real world film and camera lens tests,
as noted in their nr. 19 zeiss lens newsletter (on their website).
But I do take it as useful and interesting that David's points relate to
possible issues with scanners being unable to use lower contrast images
below some cutoff (he seems to suggest 50% as with 35 lpmm provia 100
limits). Being analog, film printing and slide projection seem to be able
to make use of high frequency data (high contrast and high resolution),
even if the contrast % falls significantly below 50% (and even 10-15%..)
This may be the difference between scanning film and printing it directly?
d) one poster says there is no significant high frequency data on the film
above 50% MTF, and the other poster says that there are "oodles" of high
frequency data encoded in the scans that aren't in the original film? Huh?
both claims don't match my understanding and experience; high resolution
slide films like provia 100 work fine above 35 lpmm and 50% contrast, and
there can't be more information after a scan than there was in the
original film being scanned, only the same at best or less in most cases.
rafe makes a very good point, which is that there are no visible
differences to him on his monitor, or on a digital print, between various
images (after decompression and other processing and manipulation, I
assume).
In my view, monitors are not really usable in comparing high resolution
imagery, for reasons of resolution limits themselves (e.g., 72 dpi
monitors..) and color mapping issues etc.
Similarly, digital prints even at 240 or 300 dpi equiv. are problematic,
partly due to the emphasis on interpolation of data in the printer
drivers, including smoothing which removes higher frequency data, and from
the relatively modest lpmm equiv. on the print (300 dpi ~5+ lpmm vs. 8
lpmm on a very good chemical print). Here again, there is a tendency to
smear color data, and the resulting prints look "creamy smooth" to me and
not "gritty" as with a chemical print ;-)
the digital cameras generally use a low pass filter to cut off higher
frequencies, which is required for anti-aliasing effects. On most
digicams, this limits resolution to circa 40 to 50+ lpmm. Given a film
camera with a high resolution film, and potential to record 70 or 80+ lpmm
per film specs and my experience, it would take a low pass filter,
essentially a "fog" filter or diffusion filter, to reduce the system
resolution to the 40 to 50+ lpmm resolution limit equivalent. Look up how
anti-aliasing filters work (Kodak etc.).
Bob Monaghan
Hi rafe and Stacey,
just to chime in that I understand tonality to be the result of more data
and step values, which is related to high frequency data content in film.
If a ten micron area of film had only 5 film grains, it would have less
data than if it had 10 film grains of smaller size. More grains provide
more info, and more info requires higher bandwidth and higher frequency
to describe and encode. The larger number of film grains can therefore
provide finer steps (and more of them) in transitions, which is the
essence of tonality ;-)
this is similar to the greater tonality in medium format vs. 35mm; the
6x6cm image would have 4 times the film area and four times the info of
the 35mm shot, for same subject height on a given enlargement (10x10").
this is why I wonder at any film scan in which the tonality is worse than
a digicam image, given the original film is rated at higher megapixel
equivalents (e.g., 40 MP). It would be like finding provia 100 used in a
35mm camera had better tonality than provia 100 used with similar quality
lenses in medium format. The degree of difference should also be the same,
i.e., 4X, because 6x6cm area is 4X that of 35mm, and provia 100's
Megapixel equiv. of 40 MP is at least 4X that of the better digital
camera generated images. Again, this is another way to say that there
should be 4X the information in the film scan. The fact that there never
is, per David's observations and rafe's experience, suggests to me that
there is something fundamental limiting film scanner resolution, such as
Nyquist response or contrast sensitivity and so on...
David J. Littleboy
"Bob Monaghan" <rmon...@engr.smu.edu> wrote:
> a) JPEG has a loss-less mode option, which provides relatively little
> compression, but enables you to recreate the original file without
> losses; and it has long been part of the full JPEG standard
My understanding it that that is only available to users with the advent of
JPEG2000.
> b) it is correct that much of the color space is being compressed,
> throwing away color data to save space. This is one way in which original
> photos (and 48 bit scans) should be superior to digital prints in color
> fidelity.
If you look at the bits, low compression jpeg reproduces the vast majority
of the pixels exactly: it is lossless for all practical purposes.
> c) despite the claim of insignificant contrast above 35 lpmm on film,
> reports of test results by lens testing services routinely report lens
> high aerial contrast (as in 60 to 80%+) and high resolutions, centrally as
> high as 450 to 600+ lpmm aerial resolution (E.g., pentax 50mm and leica
> 50mm respectively). System resolution calculations produce resolutions
> over 100 lpmm, which is what we observe in actual tests. Zeiss tests at
> 1996 photokina exceeded 200 lpmm in real world film and camera lens tests,
> as noted in their nr. 19 zeiss lens newsletter (on their website).
Real photographers don't see that in real images. Creating and illuminating
test targets itself is difficult. Those targets just don't happen in nature.
> But I do take it as useful and interesting that David's points relate to
> possible issues with scanners being unable to use lower contrast images
> below some cutoff (he seems to suggest 50% as with 35 lpmm provia 100
> limits).
That was never one of my points: I don't see anything with my microscope
that the scanner isn't getting. (If you look at K25 test target images under
a microscope you will see stuff scanners don't get.)
Provia being 50% MTF at 35 lp/mm is from:
http://www.fujifilm.com/JSP/fuji/epartners/bin/Provia100f.pdf
Negative films do better than that, but at a cost of higher grain noise.
Higher noise means lower dynamic range (in the engineering sense of the
term), and also means that high frequency detail gets swamped sooner.
> Being analog, film printing and slide projection seem to be able
> to make use of high frequency data (high contrast and high resolution),
> even if the contrast % falls significantly below 50% (and even 10-15%..)
> This may be the difference between scanning film and printing it directly?
What difference? Most people report that the better scanning technologies do
a better job of getting detail from film to the print than do projection/wet
chemistry technologies.
> In my view, monitors are not really usable in comparing high resolution
> imagery, for reasons of resolution limits themselves (e.g., 72 dpi
> monitors..) and color mapping issues etc.
Huh? You can display parts of images at any magnification you please.
> the digital cameras generally use a low pass filter to cut off higher
> frequencies, which is required for anti-aliasing effects. On most
> digicams, this limits resolution to circa 40 to 50+ lpmm. Given a film
> camera with a high resolution film, and potential to record 70 or 80+ lpmm
> per film specs and my experience, it would take a low pass filter,
> essentially a "fog" filter or diffusion filter, to reduce the system
> resolution to the 40 to 50+ lpmm resolution limit equivalent. Look up how
> anti-aliasing filters work (Kodak etc.).
This again ignores the point that when people actually look at the real life
images produced from the actual things they want to photograph, they find
that the 1Ds records and gets onto prints more detail than 35mm does, and
that the difference between 6MP dSLR images and 35mm images is so small as
to be non-existant.
That is: in dSLR megapixel terms, 35mm is slightly more than 6 and a lot
less than 12 MP.
You need a theory that explains this reality, not one that denies it.
Rafe B.
That's 0.25 INCH by 0.25 INCH.
And it's explained in several places in the text and
specifically for every snippet posted.
In a few cases folks supplied snippets of some
other size. Best I could do in those cases was
at least to be clear about the film area represented
and the native scan resolution used.
In all of my own scan snippets I used the
reference area of 0.25 inch x 0.25 inch.
>2)
>On the Howtek 4500 Drum scan vs. LS-8000
>Two Scans of the same image
>Photo and scans by Lawrence Smith
>The Overview 35 mm, Kodak 100sw pushed two stops, the highlighted areas are not
>1/4 or 1/6 of the total image.
>
>3)
>On the Heidelberg Tango Drum vs LS-8000
>Two Scans of the same image
>Photo and scans by Bill Hilton
>The Overview 645, Velvia, the boxed image is not 1/4 or 1/6 of the entire image.
>
>Giving these 3 examples, can you see why I am confused when you say "How much
>image detail from 0.25" by 0.25" of film?" which would be either 1/4 or 1/6 of
>the entire image when none of the images are 1/4 or 1/6 of the image.
Again: That's 0.25 INCH by 0.25 INCH.
If you go down to the bottom of the pag where
I discuss the digicam snippets you'll see that
I explicitly equate that to 1/24 the area of a
35 mm frame.
rafe b.
http://www.terrapinphoto.com
stacey
>
> this is why I wonder at any film scan in which the tonality is worse than
> a digicam image, given the original film is rated at higher megapixel
> equivalents (e.g., 40 MP). It would be like finding provia 100 used in a
> 35mm camera had better tonality than provia 100 used with similar quality
> lenses in medium format.
And I think many 35mm shooters are obcessed with -sharpness- and fail to see
the other things needed for technical "quality". Sure in an 8X10 you aren't
going to see a big difference in sharpness from 35mm to 6X6 but the
tonality and general image quality is obviously better. I also think if
certain files have to be sharpened or the contrast adjusted more in
software than others, that is going to affect the outcome as well.
What I've argued with Rafe is: I've seen -high end- digicam and film scans
printed and personally don't think they are as good as a -high end- optical
print from film. I don't think comparing a digicam to film using a scan of
film (and compressing it to "web quality") is worth comparing. All it tells
me is if a 6MP digicam looks better than a film scan, then the film
scanners have a problem!
--
Stacey
stacey
>
> Giving these 3 examples, can you see why I am confused when you say "How
> much image detail from 0.25" by 0.25" of film?" which would be either 1/4
> or 1/6 of the entire image when none of the images are 1/4 or 1/6 of the
> image.
I don't follow the logic here either, unless people are now printing 35mm
camera sized crops from their 4X5 film! This is like taking a 35mm camera
sized crop of 4X5 and using that to compare 4X5 to 35mm.
Why not use 1/4 of the entire film/sensor and down size them to be the same
for comparing? Of is Rafe concerned the results might be different than he
wants them to be?
--
Stacey
stacey
> On Thu, 18 Dec 2003 23:55:23 -0500, stacey <foto...@yahoo.com> wrote:
>
>>Rafe B. wrote:
>>
>>> On Thu, 18 Dec 2003 23:30:08 -0500, stacey <foto...@yahoo.com> wrote:
>>>
>>>
>>>>> The JPG's got a bit more contrast is all,
>>>>
>>>>A bit? Look at the brown wall if front of the buiulding In the tiff,
>>>>it's pretty smooth, in the jpg it's a noisy mess!
>>>
>>> It's grain, Stacey. All film has it. This is from
>>> a chrome, albeit a very old one.
>>
>>I've looked at my chromes closely and it's funny, they don't have
>>multicolored "grain" like these scans you post do and claim this noise is
>>"grain".
>
>
> Show me a scan of a print from one of your
> chromes representing the same image area:
> 0.0625 square inches of film. Then we'll talk.
I've got a 20 X 30 optical print from a 35mm frame of a bike race in my
living room shot on 400asa color negative film. It has lots of grain of
course, plenty of blank sky and pavement, and none of the grain looks
multicolored, even up -real- close. If all scans do this, optical printing
doesn't!
I project my slides on a 4 x 6 -foot- screen and have never seen this
multicolored grain either. Maybe all *scans* do this but the grain viewed
optically isn't multicolored, the multicolored garbage is digital noise
from the scan.
>
>
>>> Your only point is tonality. Yes, the TIF was
>>> pulled later -- in the context of bobm's complaints
>>> about detail -- and with slightly less contrast
>>> than the JPG. I've since fixed that; the JPG
>>> and TIF are now from exactly the same snip.
>>
>>And "changing" them makes for a equal comparison??
>
>
> If we're talking about high frequency content or
> spatial detail, it's absolutely insignificant.
>
But it totally changed the -quality- of the image.
--
Stacey
Rafe B.
>Rafe B. wrote:
>
>> On Thu, 18 Dec 2003 23:55:23 -0500, stacey <foto...@yahoo.com> wrote:
>>
>>>Rafe B. wrote:
>>>
>>>> On Thu, 18 Dec 2003 23:30:08 -0500, stacey <foto...@yahoo.com> wrote:
>>>>
>>>>
>>>>>> The JPG's got a bit more contrast is all,
>>>>>
>>>>>A bit? Look at the brown wall if front of the buiulding In the tiff,
>>>>>it's pretty smooth, in the jpg it's a noisy mess!
>>>>
>>>> It's grain, Stacey. All film has it. This is from
>>>> a chrome, albeit a very old one.
>>>
>>>I've looked at my chromes closely and it's funny, they don't have
>>>multicolored "grain" like these scans you post do and claim this noise is
>>>"grain".
>>
>>
>> Show me a scan of a print from one of your
>> chromes representing the same image area:
>> 0.0625 square inches of film. Then we'll talk.
>
>I've got a 20 X 30 optical print from a 35mm frame of a bike race in my
>living room shot on 400asa color negative film. It has lots of grain of
>course, plenty of blank sky and pavement, and none of the grain looks
>multicolored, even up -real- close. If all scans do this, optical printing
>doesn't!
20 x 30" from 35 mm still doesn't match the
magnification you're gettting from seeing these
snippets on a 17" or 22" monitor. Even at
1200 x 1600 on a 17" monitor, you're looking
at around 40x enlargement. If your screen res
is lower, the enlargment is even higher.
And your print is probably soft which is why
the grain doesn't show.
Show me a scan of a print of the equivalent
area. Till then, you have no case.
rafe b.
http://www.terrapinphoto.com
Rafe B.
>What I've argued with Rafe is: I've seen -high end- digicam and film scans
>printed and personally don't think they are as good as a -high end- optical
>print from film. I don't think comparing a digicam to film using a scan of
>film (and compressing it to "web quality") is worth comparing. All it tells
>me is if a 6MP digicam looks better than a film scan, then the film
>scanners have a problem!
The page I've put up isn't about digcams vs. film scanning.
It's about how much image detail we can coax out of 0.0625
square inch of film, by whatever means.
If you think you can beat any of the contributions so far, go
for it. Your complaints about JPG compression affecting
the results is groundless and moot.
You've got some really big prints and a flatbed scanner -
let's see you put your pixels where your mouth is.
rafe b.
http://www.terrapinphoto.com
scan comparisons
http://www.terrapinphoto.com/jmdavis
Rafe B.
wrote:
TIFFs play no meaningful role on today's Web.
I'll bet 99.99% of all web users have never seen or
been exposed to a TIF file while surfing the web.
TIFF compression (or any lossless compression
of images) can't begin to compete with JPG, even
with high-quality (ie., low-compression) JPG settings.
Lossless compression of images might save you
15-20%. Even high-quality JPG starts with at least
a 60-70% file size reduction.
rafe b.
http://www.terrapinphoto.com
Karl Winkler
Rafe,
I've loaded an "overview" image and a full-rez detail image (696x477)
of this photograph, taken with a Hassy 503cw and 120mm lens on
Ektachrome Pro (std) film. It was scanned using a Microtek 120tf at
4000ppi.
http://pages.cthome.net/karlwinkler/BlueLight.jpg
http://pages.cthome.net/karlwinkler/BlueLightDetail.jpg
Regards,
-Karl Winkler
jjs
kwin...@sennheiserusa.com (Karl Winkler) wrote:
\
> I've loaded an "overview" image and a full-rez detail image (696x477)
> of this photograph, taken with a Hassy 503cw and 120mm lens on
> Ektachrome Pro (std) film. It was scanned using a Microtek 120tf at
> 4000ppi.
>
> http://pages.cthome.net/karlwinkler/BlueLight.jpg
> http://pages.cthome.net/karlwinkler/BlueLightDetail.jpg
I don't see the accutance I'd like to, however that's not a good subject
for a rez test. The only thing one can target is the line between the blue
light and the chrome housing. BTW, what was your shutter speed? F-stop?
Rafe B.
Winkler) wrote:
Thanks, Karl.
I'll have it posted soon.
If you'd like to post or email a larger detail snippet
please send it my way.
PS: If the original is from a Hassy, why isn't the overview
image square? Are there 6x7 Hasselblads? Or is there
a bit of cropping here?
rafe b.
http://www.terrapinphoto.com
scan samples
http://www.terrapinphoto.com/jmdavis
Rafe B.
wrote:
So instead of criticizing those who choose to participate,
why not send me some scan samples to post?
Michael J Davis
Frankly I get sick of TIF files and having "this program does not
recognise this file type" because there are so many varieties of TIFF.
And these programs include Quicktime, MSpaint, MS Media player,
Photoshop, Irfanview, Netscape, etc, etc.
Mike
[The reply-to address is valid for 30 days from this posting]
--
Michael J Davis
<><
Some newsgroup contributors appear to have confused
the meaning of "discussion" with "digression".
<><
David J Taylor
> recognise this file type" because there are so many varieties of TIFF.
TIFF: Thousand Incompatible File Formats....
David
jjs
<rafe....@verizon.net> wrote:
> So instead of criticizing those who choose to participate,
> why not send me some scan samples to post?
Criticism in the form of supportive metrics is not a bad thing, Rafe. Get
off your friggin high-horse. Send me a good scanner and I'll participate -
maybe, but I don't see this as a properly conducted effort.
The way to do it is to use the very same film on all the scanners. That
will obviate a lot of variables. Then the circumstances have to be
carefully regimented (circumstances - everything in the scanning technique
from temp, to power to the machine, to focus, software settings, versions,
bus-speed, the works.) Such a method is callled 'scientific' and you
grain-sniffing wishfull thinking amateurs are in dire need of such
discipline.
Did I say Fuck Off in too many words for you to understand?
jjs
jo...@xyzzy.stafford.net (jjs) wrote:
> [...]
> The way to do it is to use the very same film on all the scanners.
And by that I mean exactly: The Very Same Piece of Film. It will have to
be passed from one tester to the next.
Rafe B.
wrote:
>Did I say Fuck Off in too many words for you to understand?
Way too many, considering that you supposedly
killfiled me months ago.
rafe b.
http://www.terrapinphoto.com
Rafe B.
wrote:
So go organize it and let us know when it's done.
Till then, your criticism is just a cheap shot.
You've contributed diddly squat.
jjs
<rafe....@verizon.net> wrote:
> Till then, your criticism is just a cheap shot.
> You've contributed diddly squat.
Oh it is clear that I added something to the effort. But continue with
your play-science. I won't disturb you again.
Karl Winkler
> On 20 Dec 2003 08:59:14 -0800, kwin...@sennheiserusa.com (Karl
> Winkler) wrote:
>
> >Rafe B. <rafe....@verizon.net> wrote in message news:<lk7ttvsu6c8an6ld0...@4ax.com>...
> >
> >Rafe,
> >
> >I've loaded an "overview" image and a full-rez detail image (696x477)
> >of this photograph, taken with a Hassy 503cw and 120mm lens on
> >Ektachrome Pro (std) film. It was scanned using a Microtek 120tf at
> >4000ppi.
> >
> >http://pages.cthome.net/karlwinkler/BlueLight.jpg
> >http://pages.cthome.net/karlwinkler/BlueLightDetail.jpg
>
>
> Thanks, Karl.
>
> I'll have it posted soon.
>
> If you'd like to post or email a larger detail snippet
> please send it my way.
>
> PS: If the original is from a Hassy, why isn't the overview
> image square? Are there 6x7 Hasselblads? Or is there
> a bit of cropping here?
>
I've got another photo that I plan to scan and include as well, from
the same day but of a different car. And with that one I'll feel more
comfortable with the full frame <g>. For both, I'll send (or post)
1000 x 1000 pixel detail shots, also. This was a quick response to
your call for entries, and I didn't yet have the time to prepare the
"proper" files.
Thanks, by the way, for putting together this (albiet subjective)
collection of scanned images. I looked through a few of the samples
and found it interesting, already.
Regards,
-Karl Winkler
http://pages.cthomet.net/karlwinkler
Stanley Krute
> I've assembled a collection of film scan samples at
>
> <http://www.terrapinphoto.com/jmdavis/>
This is great. Muchos thankos.
-- stan
Ti
says...
Ok, after reading the post by BH, I understand about the film scans.
I don't understand why the digital is arbitrarily set to 1" when being compared
to other than 35mm film. Is the actual size of the digital image when brought
up in photoshop 1"? What is the actual size? If other then 1" then shouldn't
the digital image also only use 1/4" x 1/4" for comparing?
Bob Monaghan
Hi David,
The reason film scanners can not achieve more than 40 or 50 lpmm equiv.
quality is the Nyquist law, which roughly states that you have to sample
at least twice as fast as the highest frequency signal you want to
reliably detect. This is a basic law of information theory...
Consider our 4000 dpi scanner; that is 158 dots per mm, or about 79 lpmm
given two dots (black and white) per line, yes? That is equivalent to 79
cycles per mm, where a cycle is a black/white sequence. Call it 80 lpmm to
make the math easier, and this is the rate at which we are sampling the
film.
But since we are sampling at 80 lpmm, we are at best able to reliably
reconstruct an original frequency pattern of 80 lpmm /2 or 40 lpmm (or 40
cycles/mm).
This is why none of the 4,000 dpi scans are showing significant data above
40 lpmm. The Nyquist theorem is limiting us to half the sampling rate.
So the answer is that our film scanners are limited per information theory
to circa 40 to 50 lpmm for 4000-5200 dpi range scanners (true, not
interpolated dpi).
regards bobm
PS: Loss-less compression was an option in the original JPEG draft
standards I saw in an E.E. course on the subject I took by Dr. Ricomi, one
of the experts involved in drafting the JPEG standard (and MPEG) back in
mid 1990s and on SMU's faculty...
jjs
Monaghan) wrote:
> Hi David,
>
> The reason film scanners can not achieve more than 40 or 50 lpmm equiv.
> quality is the Nyquist law, which roughly states that you have to sample
> at least twice as fast as the highest frequency signal you want to
> reliably detect. This is a basic law of information theory...
What do you mean by 'fast'? Transport speed?
>[...]
> So the answer is that our film scanners are limited per information theory
> to circa 40 to 50 lpmm for 4000-5200 dpi range scanners (true, not
> interpolated dpi).
Does that include drum scanners?
Bob Monaghan
by twice as fast I mean you have to sample at least twice the frequency;
usually we sample in time (hertz or megahertz), but in the case of an
image we are sampling in space (dpi or samples per mm corresponding to
cycles per mm, or cycles per second (hertz) in frequency domains etc.
yes, it applies to drum scanners too, anywhere you are sampling and trying
to reconstruct or digitize the data in an image, or a signal, radio wave
etc. as a basic law of information theory and sampling...
because we are taking equally spaced samples we have to sample, per
Nyquist's Law, at least twice as frequently as the highest frequency (or
lpmm) which we want to reconstruct.
This is a law of info theory, so like physics it applies to all such
situations whether CCD or CMOS or drum scanner optics etc. where we are
taking samples (vs. continuous processes like analog film enlarging..)
regards bobm
Rafe B.
>Ok, after reading the post by BH, I understand about the film scans.
>
>I don't understand why the digital is arbitrarily set to 1" when being compared
>to other than 35mm film. Is the actual size of the digital image when brought
>up in photoshop 1"? What is the actual size? If other then 1" then shouldn't
>the digital image also only use 1/4" x 1/4" for comparing?
Sigh. I thought I had explained it pretty well on
the website, and in previous posts in this thread.
Must say it's getting a bit tiresome trying to
explain this.
There is no "actual size" to a film scan, nor
an "actual size" to a digicam capture. There
is only an image, X pixels wide by Y pixels high.
When I scan a 35 mm frame on an LS-8000
at 4000 dpi, I get an image roughly 4000
pixels high and 6000 pixels wide. Well,
not exactly -- it's about 10% smaller than
that in both dimensions, but let's not quibble.
When I take a shot with my Canon 10D I get
an image 2048 pixels high by 3072 pixels
wide. Exactly.
So in the latter case, you could think of the
10D as a "film scan" of a 1" by 1.5" area
at 2048 dpi. End of story and QED.
Please, Ti. Do the math. It's pretty simple -
assuming you understand what "dots per
inch" mean.
rafe b.
http://www.terrapinphoto.com
scan snippets
http://www.terrapinphoto.com/jmavis
Rafe B.
wrote:
>
>Hi David,
>
>The reason film scanners can not achieve more than 40 or 50 lpmm equiv.
>quality is the Nyquist law, which roughly states that you have to sample
>at least twice as fast as the highest frequency signal you want to
>reliably detect. This is a basic law of information theory...
>
>Consider our 4000 dpi scanner; that is 158 dots per mm, or about 79 lpmm
>given two dots (black and white) per line, yes? That is equivalent to 79
>cycles per mm, where a cycle is a black/white sequence. Call it 80 lpmm to
>make the math easier, and this is the rate at which we are sampling the
>film.
>
>But since we are sampling at 80 lpmm, we are at best able to reliably
>reconstruct an original frequency pattern of 80 lpmm /2 or 40 lpmm (or 40
>cycles/mm).
>
>This is why none of the 4,000 dpi scans are showing significant data above
>40 lpmm. The Nyquist theorem is limiting us to half the sampling rate.
>
>So the answer is that our film scanners are limited per information theory
>to circa 40 to 50 lpmm for 4000-5200 dpi range scanners (true, not
>interpolated dpi).
>
>regards bobm
As of this PM, I've posted 1600 dpi scan snippets of optical
prints -- Cibachromes at 8x10" from 35 mm Ektachromes.
Odd, but they don't show any more detail -- in fact a bit
less -- than my 4000 dpi film scans, from a print area
referenced back to 0.25" x 0.25" on the film. The Ciba
prints were scanned at 1600 dpi, which by all accounts
is quite generous for this purpose.
But of course, there are plenty of other ways to
explain the lack of resolved detail...
>PS: Loss-less compression was an option in the original JPEG draft
>standards I saw in an E.E. course on the subject I took by Dr. Ricomi, one
>of the experts involved in drafting the JPEG standard (and MPEG) back in
>mid 1990s and on SMU's faculty...
Lossless JPG exists but mostly for very special purposes,
and generally not for consumer or even professional
graphics image storage.
You can get the source code for it from the jpg.org web site.
It is used (for example) in some scanners to transmit the
scan data back to the host.
Nowadays, the preferred lossless/compressed image
format is PNG. But even the creators of PNG admit that
the available compression isn't in the same ballpark as JPG.
rafe b.
http://www.terrapinphoto.com
scan snippets
Ken Weitzel
> Sigh. I thought I had explained it pretty well on
> the website, and in previous posts in this thread.
> Must say it's getting a bit tiresome trying to
> explain this.
>
> There is no "actual size" to a film scan, nor
> an "actual size" to a digicam capture. There
> is only an image, X pixels wide by Y pixels high.
I'm just barely graduating from the confused state
myself, so perhaps if I try to explain where all my
confusion stems from some of you might be able to
re-word it more eloquently to help others get over
the "hump" :)
Let me respectfully say I strongly disagree with your
statement that there is no "actual size" of a digicam
picture. I insist there is. On the other hand, my
(13 yr old) grand daughter finds it much easier to
agree with you.
The problem stems from the fact that I started out
hiding in a dark basement with trays, chemicals, a
roll of film, a contact printer, and some paper.
If memory serves, the film was 2.25 inches. Holy smokes,
that was the actual size! So was (virtually) the paper.
The actual size!
Now, I take a pic with my Olympus. I think it has
a 2/3 inch or thereabouts sensor in it. *That's* the
actual size of the image!!! But I dl it, load it into
PSP and PSP says the image size is something like 22 inches
or something. And then PSP has the audacity to tell me
I'll have to shrink it when I try to drop it onto 6 inch
paper to print it!!!! :(
Now I'm getting old, and stroke damaged, so can't say it
well, but perhaps one of you out there who's still younger
and articulate can use this to help others who are "stuck"
like I was.
Take care.
Ken
Bill Hilton
>I take a pic with my Olympus. I think it has
>a 2/3 inch or thereabouts sensor in it. *That's* the
>actual size of the image!!
Good point.
>But I dl it, load it into PSP and PSP says the image
>size is something like 22 inches or something.
No, most likely what's happening here is that the resolution is defaulting to
72 ppi (selected because it's a typical monitor resolution) so you get this
huge size, but that's totally meaningless until you go to print it. You can
change the rez to anything you want and if you do not resample the pixels the
total pixel count doesn't change.
>And then PSP has the audacity to tell me
>I'll have to shrink it when I try to drop it onto 6 inch
>paper to print it!!!! :(
If it's like Photoshop there is an option to resize and resample, or an option
to resize without resampling, which is probably what you want. If it's 22
inches at 72 ppi then it's probably 6 inches at 264 ppi.
Bill
David J. Littleboy
"Bob Monaghan" <rmon...@engr.smu.edu> wrote in message
news:bs322g$a4p$1...@blaze.seas.smu.edu...
>
> Hi David,
>
> The reason film scanners can not achieve more than 40 or 50 lpmm equiv.
> quality is the Nyquist law, which roughly states that you have to sample
> at least twice as fast as the highest frequency signal you want to
> reliably detect. This is a basic law of information theory...
Thank you. I understand that quite well.
What I also understand is that if you make a 13x19 from 35mm, it'll look
like soft mush (whatever film-to-paper technology you use), and if you make
a 13x19 from an 11MP 1Ds image, it'll look quite a bit nicer. And if you
make a 13x19 from 645, it might look a tad nicer than the 1Ds. Maybe.
This isn't about scanning; it's about film having inadequate performance to
support 13x enlargements.
Film being inadeqiate to support 13x enlargements shouldn't be news to you:
it is, in fact, common knowledge that everyone agrees on. Even 645 is
considered inadequate to make 13x19 prints.
(Aside: Since film has inadequate performace to support 13x enargements,
4000 dpi is a nice overkill resolution to scan at since profession digital
printing uses 300 dpi as standard.)
The point here, is that the various theories you've proposed ("film
_usefully and meaningfully_ resolves 100 lp/mm, as well as your incessant
quoting the "ISO 400 35mm is worth 23MP" and "ISO 100 35mm film is worth
40MP" insanities from Kodak) are grossly at odds with objective reality.
David J. Littleboy
Tokyo, Japan
Rafe B.
wrote:
>Hi...
>
>> Sigh. I thought I had explained it pretty well on
>> the website, and in previous posts in this thread.
>> Must say it's getting a bit tiresome trying to
>> explain this.
>>
>> There is no "actual size" to a film scan, nor
>> an "actual size" to a digicam capture. There
>> is only an image, X pixels wide by Y pixels high.
>
>I'm just barely graduating from the confused state
>myself, so perhaps if I try to explain where all my
>confusion stems from some of you might be able to
>re-word it more eloquently to help others get over
>the "hump" :)
>
>Let me respectfully say I strongly disagree with your
>statement that there is no "actual size" of a digicam
>picture. I insist there is. On the other hand, my
>(13 yr old) grand daughter finds it much easier to
>agree with you.
Children have open minds and are far better
learners than adults. That's an old story.
Of course there is an actual size to the digicam
sensor, but I'm trying to make that irrelevant by
equating the digicam picture to a 35 mm filmscan.
And what's wrong with that? Hasn't the issue been,
for these last several years, whether digicams
have "caught up" to film? And when that statement
is made (right or wrong) what film is it that's being
talked about? You guessed it: 35 mm.
When we argue or chat about the "megapixel
equivalent" of film, isn't it usually 35 mm that's
taken as the reference?
*Of course* anyone with a working brain
understands that larger film formats hold more
total information. It's so obvious that it goes
without saying. Or so I would hope.
Is the digicam sensor size relevant? Yes
and no. My study is based only on the
resulting image -- the end result -- so, from
my narrow perspective here, I'll say no.
Of course it's a lie! It's a given that bigger
sensors give better images -- and that's
just as true for digicams as for film. But
I'm trying to focus on the results, not the
mechanics of the image capture.
If there is a benefit to a bigger sensor, it
should show up with my metric as well --
and it does.
>The problem stems from the fact that I started out
>hiding in a dark basement with trays, chemicals, a
>roll of film, a contact printer, and some paper.
Same here, for several years. I was printing
8x10s and 11x14s from BW film nearly forty
years ago. I have a couple of scans now on my
website from Ciba prints made approximately 25
years back.
>If memory serves, the film was 2.25 inches. Holy smokes,
>that was the actual size! So was (virtually) the paper.
>The actual size!
Well, for some it was 24 x 36 mm, for others it
was 4x5", for some of us it was even 36 x 36 mm
or 18 x 24 mm, or 6x7 cm, or... well, you get the picture.
>Now, I take a pic with my Olympus. I think it has
>a 2/3 inch or thereabouts sensor in it.
Not by a longshot. More like maybe 4 mm by 7 mm
if it's a point-and-shoot. The sensors in the 10D and
D100 class are 15 x 22 mm. But again, I'm proposing
a metric that makes the sensor size irrelevant, by
recognizing only the pixel dimensions of the resulting
digital image.
It's an arbitrary metric, but a useful one. If you
understand that 645 is, say, 3.5 x the film area of
35 mm, you should be able to make the mental
leap from my results to a 645.
>*That's* the actual size of the image!!! But I dl it, load it into
>PSP and PSP says the image size is something like 22 inches
>or something. And then PSP has the audacity to tell me
>I'll have to shrink it when I try to drop it onto 6 inch
>paper to print it!!!! :(
Shrink it? No way. There's no way you want to give
up one single pixel. The change you need to make
doesn't add, subtract, modify or rearrange the image
pixels at all.
>Now I'm getting old, and stroke damaged, so can't say it
>well, but perhaps one of you out there who's still younger
>and articulate can use this to help others who are "stuck"
>like I was.
Just try to keep an open mind, is all. Yeah, I'm getting
on in years also, but certain trends ought not be ignored.
Technologies come and go.
rafe b.
http://www.terrapinphoto.com
scan snippets
Ti
says...
> On Sun, 21 Dec 2003 02:05:02 GMT, Ti <T...@hotmail.com> wrote:
>
>
> >Ok, after reading the post by BH, I understand about the film scans.
> >
> >I don't understand why the digital is arbitrarily set to 1" when being compared
> >to other than 35mm film. Is the actual size of the digital image when brought
> >up in photoshop 1"? What is the actual size? If other then 1" then shouldn't
> >the digital image also only use 1/4" x 1/4" for comparing?
>
>
> Sigh. I thought I had explained it pretty well on
> the website, and in previous posts in this thread.
> Must say it's getting a bit tiresome trying to
> explain this.
Sorry about making this tiresome for you, but I'm not trying to be difficult.
I'm just trying to understand why?
>
> There is no "actual size" to a film scan, nor
> an "actual size" to a digicam capture. There
> is only an image, X pixels wide by Y pixels high.
>
> When I scan a 35 mm frame on an LS-8000
> at 4000 dpi, I get an image roughly 4000
> pixels high and 6000 pixels wide. Well,
> not exactly -- it's about 10% smaller than
> that in both dimensions, but let's not quibble.
>
> When I take a shot with my Canon 10D I get
> an image 2048 pixels high by 3072 pixels
> wide. Exactly.
>
> So in the latter case, you could think of the
> 10D as a "film scan" of a 1" by 1.5" area
> at 2048 dpi. End of story and QED.
>
> Please, Ti. Do the math. It's pretty simple -
> assuming you understand what "dots per
> inch" mean.
Perhaps I don't understand what dpi means because I don't equate an image that
is 2048 pixels wide as being 2048 dpi. I see pixels and dpi as different.
Although 1" was arbitrarily picked for comparison, that doesn't make it a fact
that the digital image is 1".
David J. Littleboy
"Ti" <T...@hotmail.com> wrote:
> >
> > So in the latter case, you could think of the
> > 10D as a "film scan" of a 1" by 1.5" area
> > at 2048 dpi. End of story and QED.
> >
> > Please, Ti. Do the math. It's pretty simple -
> > assuming you understand what "dots per
> > inch" mean.
>
> Perhaps I don't understand what dpi means because I don't equate an image
that
> is 2048 pixels wide as being 2048 dpi. I see pixels and dpi as different.
> Although 1" was arbitrarily picked for comparison, that doesn't make it a
fact
> that the digital image is 1".
The idea is to _pretend_ that a digital sensor is the same size as the 35mm
frame.
This is because dSLR digital is thought to be in competition with 35mm.
The 35mm frame is roughly 1" high, so it makes sense to think of a dSLR
image as being a 2048 dpi scan of a 1.0 x 1.5" frame. This underestimates
the quality of the digital image by a factor of 24/25.4. Close enough for
government work.
Rafe B.
>Perhaps I don't understand what dpi means because I don't equate an image that
>is 2048 pixels wide as being 2048 dpi. I see pixels and dpi as different.
>Although 1" was arbitrarily picked for comparison, that doesn't make it a fact
>that the digital image is 1".
It is not a *fact* nor did I ever claim it was.
It is nothing more than a convenient means of
comparison, which I have explained several
times now. Take it or leave it, it's all the same
to me.
The square root of minus one isn't a fact.
It does not and cannot exist, except in the
abstract.
Yet by allowing it to exist *in the abstract* it has
transformed the study and teaching of numerous
physical sciences and mathematics.
rafe b.
http://www.terrapinphoto.com
scan snippets
Neil Gould
> "Bob Monaghan" <rmon...@engr.smu.edu> wrote in message
> news:bs322g$a4p$1...@blaze.seas.smu.edu...
>>
>> Hi David,
>>
>> The reason film scanners can not achieve more than 40 or 50 lpmm
>> equiv. quality is the Nyquist law, which roughly states that you
>> have to sample at least twice as fast as the highest frequency
>> signal you want to reliably detect. This is a basic law of
>> information theory...
>
> Thank you. I understand that quite well.
>
> What I also understand is that if you make a 13x19 from 35mm, it'll
> look like soft mush (whatever film-to-paper technology you use), and
> if you make a 13x19 from an 11MP 1Ds image, it'll look quite a bit
> nicer. And if you make a 13x19 from 645, it might look a tad nicer
> than the 1Ds. Maybe.
>
What does "nicer" mean? In what dimension?
> This isn't about scanning; it's about film having inadequate
> performance to support 13x enlargements.
>
> Film being inadeqiate to support 13x enlargements shouldn't be news
> to you: it is, in fact, common knowledge that everyone agrees on.
> Even 645 is considered inadequate to make 13x19 prints.
>
> (Aside: Since film has inadequate performace to support 13x
> enargements, 4000 dpi is a nice overkill resolution to scan at since
> profession digital printing uses 300 dpi as standard.)
>
> The point here, is that the various theories you've proposed ("film
> _usefully and meaningfully_ resolves 100 lp/mm, as well as your
> incessant quoting the "ISO 400 35mm is worth 23MP" and "ISO 100 35mm
> film is worth 40MP" insanities from Kodak) are grossly at odds with
> objective reality.
>
The discussions that I've seen on this topic always presume something
about the image without being clear. When one compares images based on the
size of the data, I suspect that they're only looking at the apparent
"sharpness" (or pseudo-sharpness) of the image, rather than, for example,
the tonal gradation. What is being disregarded is that a 300 dpi image @
36 bits (12 b/p) *will* be quite a bit smaller than that image at 48 bits
(16 b/p). So, it makes no sense to me to make statements about the
"insanities" of the relative data sizes *unless* one refers to *both* the
xy resolution *and* the bit depth of images.
What I'm calling "pseudo-sharpness" is the higher contrast that
accompanies lower bit depths. It makes the transitions between colors
appear "sharper" simply by the inability to represent the original
gradations in an image. Consequently, images that *can* represent the
gradations may appear "softer".
A key question that addresses the relationship between data size and
images is the equivalent bit depth of "film". Perhaps Kodak and others
have a different notion of what it takes than those that think that 11 mp
is more than enough to match 35 mm film? A lot depends on the composition.
So, I really am confused when comments about "nicer" images are made in
lieu of comments about the nature of the image.
Another presumption is that there is some relevance to 300 dpi. This
became a popular notion when creating images for offset presswork about 20
years ago, and later carried over to inkjet technologies. Prior to 20
years ago, the limitation of pre-press to reproduce xy resolution was
based on the availability of separation screens. As scanning technology
and imagesetters replaced mechanical separations, the xy resolution has
continuously increased over time. Therefore, I suspect that these
limitations are matters of convenience based on the notion that if a
printer can't resolve images above 300 dpi, then anything more than that
is irrelevant. Now that it is commonplace for offset presswork to print
beyond 150 lpi (the original basis for 300 dpi resolution based on the
Nyquist theorem), the ideal xy resolution has increased accordingly.
Similarly, as Hexachrome technologies can successfully represent a wider
gamut than CMYK, the greater bit depth of images could be exploited. A 200
lpi offset job is not the exotic dream that it was 20 years ago, and 400
dpi images are the "norm" for that kind of work. Even higher line screens
are used in today's "exotic" presswork. Coupled with Hexachrome, the
requisite data size of an acceptable source file has grown considerably.
As inkjet technology continues to improve, both through the use of higher
xy resolution and multiple ink colors for increasing the bit depth
representation, I expect to see the same kind of increase in the data size
of "acceptable" source files.
Regards,
--
Neil Gould
--------------------------------------
Terra Tu AV - www.terratu.com
Technical Graphics & Media
jjs
Monaghan) wrote:
> by twice as fast I mean you have to sample at least twice the frequency;
> usually we sample in time (hertz or megahertz), but in the case of an
> image we are sampling in space (dpi or samples per mm corresponding to
> cycles per mm, or cycles per second (hertz) in frequency domains etc.
> [... snip good article ...]
So _very_ well put. Thank you. I get it now.
jjs
<dav...@gol.com> wrote:
> [...]
> (Aside: Since film has inadequate performace to support 13x enargements,
> 4000 dpi is a nice overkill resolution to scan at since profession digital
> printing uses 300 dpi as standard.)
Whew, strong words but given that until more people finally had a
'4000spi' scanner, the 'answer' was drum scanning and people were making
what now seem to be extraordinary claims about drumming. Have we finally
found the answer and come to the watershed?
jjs
Gould" <ne...@myplaceofwork.com> wrote:
> The discussions that I've seen on this topic always presume something
> about the image without being clear. When one compares images based on the
> size of the data, I suspect that they're only looking at the apparent
> "sharpness" (or pseudo-sharpness) of the image, rather than, for example,
> the tonal gradation. What is being disregarded is that a 300 dpi image @
> 36 bits (12 b/p) *will* be quite a bit smaller than that image at 48 bits
> (16 b/p). So, it makes no sense to me to make statements about the
> "insanities" of the relative data sizes *unless* one refers to *both* the
> xy resolution *and* the bit depth of images.
>
> What I'm calling "pseudo-sharpness" is the higher contrast that
> accompanies lower bit depths. It makes the transitions between colors
> appear "sharper" simply by the inability to represent the original
> gradations in an image. Consequently, images that *can* represent the
> gradations may appear "softer".
> [... great, just great article ...]
I will stop responding after this, but this post, Neil's contribution,
does rather fracture the previous arguments. It seems were were assuming
the equivalent of a Flat Earth! Resolution - what I call 'organic' forms
(real life forms of smooth convolutions) compared to military resolution
targets is a huge challenge requiring image-information depth. Maybe the
rez boyz should be given a new target: illuminated by two diffused light
sources; a still-life of white eggs floating in rainbow jelly with a
cotton fluff background all balanced on the tummy of a marbled venus.)
Pardon me, but I have the killer flu and it's the fever. That's my excuse
and I'm sticking to it.
> Another presumption is that there is some relevance to 300 dpi.
> [... and it gets better and better ...]
IMHO this particular section (date,time) of this whole thread should be
published in a consumer periodical for the rest of us.
Karl Winkler
Here's one of the two I had in mind:
http://pages.cthome.net/karlwinkler/FlamesBsm.jpg
http://pages.cthome.net/karlwinkler/FlamesBdetail.jpg
These were shot with a Hasselblad 503cw and 120mm CFi lens on Fuji
Astia 100 film. I used the Microteck Artixscan 120tf at 4000 ppi. For
the "overview" full frame shot, I dithered down the image in
Photoshop. The detail shot is a straight, full size piece, at 1000
x1000 pixels.
Regards,
-Karl
Rafe B.
Winkler) wrote:
Excellent. Many thanks. I'll have these
posted shortly.
Alan Browne
Rafe B. wrote:
>
>
> Children have open minds and are far better
> learners than adults. That's an old story.
Double edged sword. Primacy is the first law of teaching, and if
children are taught the wrong thing first, it tends to stick harder than
the correction. Religions naturally take full advantage of primacy.
Alan Browne
Ken Weitzel wrote:
>
> Now, I take a pic with my Olympus. I think it has
> a 2/3 inch or thereabouts sensor in it. *That's* the
> actual size of the image!!! But I dl it, load it into
> PSP and PSP says the image size is something like 22 inches
> or something. And then PSP has the audacity to tell me
> I'll have to shrink it when I try to drop it onto 6 inch
> paper to print it!!!! :(
Yes, the "actual size" at capture is the size of the sensor.
However, pixels in storage have no fixed dimension.
They 'inflate' to a huge looking size on your screen ('cause your screen
is only 85 dpi (or so) and several times smaller than that on your
printer (at 300 dpi) ...still much larger than the 3600 dpi of the
Olympus E-1 sensor. (assuming that's the sensor you have...).
PSP uses default printer sizes. The default size you have makes that
image 22 inches... The default can be anything... But preparing your
image for printing on a 300 dpi printer may require resizing the image,
and discarding information from the original capture. 300 dpi * 6
inches requires a mere 1800 pixels of width, after all...and your camera
caught more than that ... (also keep in mind whatever cropping you
might do).
So, keep your "original" copy of the image at full size (in pixels),
then adjust size for the print, redo the USM at this point on the
re-sized image, and print away.
It's really no different than a negative in an enlarger. The original
35mm is 36x24mm regardless. The enlarger allows scaling as required to
any size...(well...)
Cheers,
Alan
Wayne Fulton
rafe....@verizon.net says...
>
>It is not a *fact* nor did I ever claim it was.
It is a very interesting page, and I am glad that you put it up, thank you.
However, FWIW, I dont get the 1 inch either. It is an imaginary random number
without meaning or basis in fact.
If you are going to compare to 1/24 of the film area, then why not just
compare with 1/24 of the digital camera pixels too? That would at least be
equally proportional to capability to print an image.
--
Wayne
http://www.scantips.com "A few scanning tips"
Alan Browne
Rafe B. wrote:
>>>Rafe B. wrote:
>>>
>>>
>>>>I beg to differ. There is no observable difference in this
>>>>case between the JPG and the TIFF. No details have
>>>>been lost. Furthermore, TIFFs can't be viewed by web
>>>>browers and the file sizes of these images would be huge.
>>>
>>>
>>>Netscape 7 accesses TIF's via a viewer such as Quicktime.
>>
>>And TIFFs can be compressed.
The point of my (and jjs') reply was not to assert that TIFF's are more
appropriate, rather simply to say that TIFF's CAN be viewed web browsers
and (per jjs) can be compressed. You made a very B&W statement that was
erroneous in two ways.
--
e-meil: there's no such thing as a FreeLunch.
Alan Browne
Michael J Davis wrote:
> And these programs include Quicktime, MSpaint, MS Media player,
> Photoshop, Irfanview, Netscape, etc, etc.
Funny ... Netscape uses Quicktime to access TIFF on my system ... and
Photoshop DEFINITELY reads and writes TIFF. I've never had an
unreadable TIFF file cross my machine.
Cheers,
Alan.
Rafe B.
wrote:
And that is *exactly* what the one inch rule is for!
Let's presume that a 35 mm frame is 1 inch
by 1.5 inch. That's a pretty close approximation
to reality and yields nice easy numbers to work with.
OK, do we agree that 0.25 inch by 0.25 inch
is 1/24 of the area of a 35 mm frame? Here's
the math:
1 x 1.5 = 1.5 (area of 35 mm frame in sq. inch.)
0.25 x 0.25 = 0.0625 (area of scan snip in sq. in.)
0.0625 / 1.5 = 0.041667 = 1/24
OK., now for the digicam. Let's pick the
10D for example. Its native image size
at full resolution is 2048 x 3072 pixels.
What square array of pixels from the 10D
represents 1/24 of the 10D's available pixels?
Let's see:
2048 * 3072 = 6291456 (approx. six million)
6291456 / 24 = 262144 (= 2^20)
sq. root (262144) = 512 (= 2^10)
Or... we could have figured this by my simpler
one inch rule: Think of the 10D as a filmscanner
at 2048 dpi -- ie. by "taking its height as one inch."
So a 1/4" by 1/4" snippet is 512 pixels on a side.
********* That's all it is, folks *********
It's a dirt simple math shortcut to figure the digicam
snippet that's "equivalent" to 0.25" x 0.25" of film.
Bob Monaghan
The more important question is:
Are current scanners (and printers) good enough?
My suspicion is that current scanners (4000/5600 dpi range) are "good
enough", meaning that their apparent limit of achieving ~40-50+ lpmm from
scanning film matches the typical techniques of most shooters (see
discussion http://medfmt.8k.com/mf/limits.html
Second, the image info available at 40 lpmm is sufficient to produce an
acceptable quality image when printed at 300 dpi (5 lpmm on print). Most
folks probably rarely got this quality from minilab printers in past etc
Third, film captures more high frequency data which produces the high
contrast and high resolution in high quality chemical prints. Being analog
and directly printed, even relatively low contrast but high frequency data
captured on film can be recorded in the slide or printed images.
It appears from David L. et. al.'s experience that contrast levels around
50% or so (~ 35 lpmm on provia 100) are needed for good quality scans. It
also appears that some interactions of scanners with certain films enhance
the apparent grain "noise" above what would be expected.
Some costly high end (drum laser..) scanners with very high real dpi
levels may be able to recover rather more high frequency data from film,
depending on their ability to work with lower contrast but high frequency
data on film (e.g., 100 lpmm range). Using a laser and photomultiplier
setup may also obviate some losses in optical lenses and contrast limits
etc. used in lower cost prosumer scanners.
High quality (8 lpmm) chemical high silver content prints and laser
generated cibachromes/R prints may be technically superior to current
inkjet printer images (e.g., at 300 dpi), but again, the inkjet printers
are "good enough", and few people have high quality chemical prints for
comparison.
Given the growing popularity of direct capture at lower costs, my final
projection is that we may NOT see improvements in film scanners for
consumers, precisely because they are "good enough" now. ;-)
Who is going to invest in bringing out better film scanners, when direct
capture is seen as equal to film at costs now less than the prosumer film
scanners now in the pipeline?
Similarly, if 50 lpmm is the limit for 99% of the images taken because of
poor technique or film choices, then where is the market (in that 1%) to
justify the high costs of developing a high end prosumer scanner for a
small and declining market?
To relate to the RPE-MF newsgroup, the followup point is that higher image
quality from current prosumer scanners is best achieved by larger film
negative sizes (cf David L's 645 experiences). So the benefits from
upgrading from 35mm to MF may carry over to folks planning on scanning
film and printing at home, as well as those doing darkroom or pro lab
prints...
grins bobm
Rafe B.
<alan....@FreeLunchVideotron.ca> wrote:
>The point of my (and jjs') reply was not to assert that TIFF's are more
>appropriate, rather simply to say that TIFF's CAN be viewed web browsers
>and (per jjs) can be compressed. You made a very B&W statement that was
>erroneous in two ways.
Up until bob m made his assertion about JPG losses,
I don't believe I had ever dealt with a TIF file on the
web. The web is dominated by JPGs and GIFs.
Internet Explorer 6.0.28 -- by itself -- cannot read TIFs.
Ditto for Netscape 7.0. Of course, there are any number
of plugins that will *add* that capability. I never had need
of that until bob m's challenge.
Furthermore there is absolutely no comparing the
degree of compression from JPG with that of any
lossless image compression format.
All of which is rather beside the point. As it is,
the images on my site are **huge** even as low-
compression JPGs. It would have been utterly
ridiculous to use TIFFs instead.
For the doubters, I've posted and given a link
(on the site) to one of the many scan snippets
as a TIFF. To my eyes the two are virtually
identical, particularly in their ability to render
the original details from a high-resolution 4000
dpi film scan.
jjs
<rafe....@verizon.net> wrote:
> All of which is rather beside the point. As it is,
> the images on my site are **huge** even as low-
> compression JPGs. It would have been utterly
> ridiculous to use TIFFs instead.
It is about samples, right? So cut out the critical section and save as
lossless JPG. You can save as virtually losselss. That is FAPP lossless.
Or better, TIFF. You only work with sections so that's a no brainer.
Concerning the overall image you want to show, just make another scan to
the web-target size, or use the last resort of the scoundrel and
downsample.
Alternately - I was serving lossless gigabit images over a year ago. I
proposed them for people who wanted to see an image in their browser, and
optionally zoom into see the lossless data. Nobody was interested. Now you
Photoshop Version 7 or CS people can do it yourself. It's built-in. Feel
free. (And there are other options which do not require plug-ins.) This is
my business, gentlemen. I do it every day. But you got to get over this
sampling standard confusion. That's your affair. So far it's not going
well for you.
jjs
Monaghan) wrote:
> [...]
> Second, the image info available at 40 lpmm is sufficient to produce an
> acceptable quality image when printed at 300 dpi (5 lpmm on print). Most
> folks probably rarely got this quality from minilab printers in past etc
Do the math for us innumerates. If I scan a (nominal) normal 58cm square
B&W negative at 4000spi, at what enlargement will it peter out? Or, which
will do better for a 20" square print, a scan-to-digital-to-digital print
or a conventional enlarger with a 100 lp/mm lens?
> [...]
David J. Littleboy
"jjs" <jo...@xyzzy.stafford.net> wrote in message
news:john-21120...@ip-0-251.sprint-rev.hbci.com...
> In article <bs597v$t6u$1...@blaze.seas.smu.edu>, rmon...@engr.smu.edu (Bob
> Monaghan) wrote:
>
> > [...]
> > Second, the image info available at 40 lpmm is sufficient to produce an
> > acceptable quality image when printed at 300 dpi (5 lpmm on print). Most
> > folks probably rarely got this quality from minilab printers in past etc
>
> Do the math for us innumerates.
The number of pixels you get is (horizontal inches x dpi) x (vertical inches
x dpi). For 6x6, that's (2.2 x 4000) x (2.2 x 4000) = 77.44 MP. Since no one
has ever seen a 4000 dpi scan that looks as good at the pixel level as a
mediocre digital camera original, Bobm concludes that the scanners are
faulty and I conclude that there isn't that much information on the film. My
microscope agrees with me and my scanner. Bobm uses a different microscope,
I guess.
> If I scan a (nominal) normal 58cm square
> B&W negative at 4000spi, at what enlargement will it peter out? Or, which
> will do better for a 20" square print, a scan-to-digital-to-digital print
> or a conventional enlarger with a 100 lp/mm lens?
Rafe has scans of both a high-quality print (1600 dpi from the print) and
4000 dpi from the film. You can't tell the difference.
The experience here is that from a 4000 dpi scan, 8x enlargements (500 dpi
from 4000 dpi scans) look very good and 13x enlargements look painfully soft
(300 dpi from 4000 dpi scans). (Observing 8.25x11.5" prints that correspond
to crops.) My experience, my microscope, and everything I've read, indicates
that that's par for the course for film. But Bobm seems to think film holds
up to much higher enlargements than that.
Wayne Fulton
rafe....@verizon.net says...
>
>And that is *exactly* what the one inch rule is for!
No, the one inch is only imagined, and it plays no part in any reality.
Yes, 512x512 pixels is 1/24 of the total 3072x2048 pixels
(however the dummy one inch argument sheds no light on that fact <g>)
I dont question any aspect of the 1/24 value or goal.
35 mm film is not 1.5x1 inch of course, but is instead at most 36x24 mm, with
the actual usually being more like 1.4x0.9 inches. (my point is that the one
inch notion fails here too)
But I dont understand the need to distract from your good work with a
notion of a one inch dimension that is not real, and has no relevance. It is
unnecessary to imagine the camera image height is 1 inch (it has no such
dimension). You simply made that part up, and you should realize now that it
detracts badly, for no good purpose.
If you use the 1.5x1.0 inch film numbers, then obviously
W/6 is 1.5 inches / 6 = 0.25 inch
H/4 is 1.0 inch / 4 = 0.25 inch
W/6 and H/4 also works for 3072x2048 pixels, and is equal to 512x512 pixels.
This seems very clear that both concepts represent 1/24 of the total area.
It seems a fair comparison of capability.
Why add the confusing one inch diversion that has no relevance ?
Again, I like the page you have created, I think it's great. I have no
problem with any aspect of your results, comparisons, method, fairness,
accuracy, approximations, numbers, or purpose, with the one exception that
the one inch notion is simply misleading and distracting noise that has no
purpose or relevance to the rest. I'm only asking about your documentation,
not the results. The digital image simply has no one inch dimension, and it
only causes confusion when you say "imagine it does". It doesnt. It would be
much stronger page (no bickering about the details) if you simply deleted the
one inch statement. It is a weakness which you dont need to make any point.
Bob Monaghan
David L. gets a value of ~35 lpmm max. from provia 100 when scanned with
very good technique and prosumer gear. As he and rafe and others have
noted, prosumer film scanners don't seem to be able to get much beyond 50
lpmm from scanning images. If these are the scanners that you use, you
won't get the same degree of image quality and enlargeability using
scanned film as you would from direct printing, IMHO. After all, provia is
rated at 140 lpmm (1000:1 hi contrast ratio) and 60 lpmm (1.6:1 low
contrast), roughly 100 lpmm for typical 6:1 contrast ratio in many images.
Even 60 lpmm is waay above 35 lpmm, right? ;-) So if all things were
equal, then you would expect at least double (60 vs. 35 lpmm) and maybe
triple (100 lpmm at 6:1 vs. 35 lpmm scanner) the enlargement factor from
using film directly in enlarger...
all things aren't equal ;-) the printers interpolate or smooth the data
they do have (at 35 lpmm), and so you can create larger enlargements with
the smoothed data than the raw 35 lpmm value would suggest. Unfortunately,
you can't add back in the missing high frequency data (above 35 lpmm to
140 lpmm with provia 100, for example) lost during scanning. We know this
data is there because the film mfger and real world lab tests have rated
the films with the above value for provia 100 etc.
the result is that you can probably get similar sized enlargements; the
digital prints will be "creamy smooth" from interpolation, with less
high contrast and high resolution ("gritty") than the chemical prints.
The scanner also reduces the color depth from the original slide (48 bits
or more), and the various algorithms modify and compress the color space
(e.g., JPEG etc.). In general, the eye isn't very good at detecting these
shifts, without a comparison print side by side. Because colors are
smoothed, elements like the sky may be more homogeneous in a digital print
than they would be in a chemical print, though the chemical print would be
the more accurate recorder of actual scene colors and brightness (being
analog).
So you pays your money and you takes your choice ;-) Technically, a 5 lpmm
digital print is not as high quality as younger eyes can see (that's 8
lpmm), but probably as good as us oldsters past 30 can use? ;-)
For me, this is less of an issue, as I prefer to shoot slides, where the
weaknesses of digital are readily apparent and film's superiority obvious
pioe[rmv]
> Given the small amount of high frequency
> information (other than grain noise) in scans, the enormous tiff file macho
> in scanner users is quite misplaced.
No, this is positively wrong. If you want the highest possible quality
and editability, everything must be recorded and stored in lossless
formats. Not only are you losing detail, JPEG means you are limited to 8
bits per channel.
You are supposed to know this, David, and you should not encourage the
kind of cutting corners it is to use JPEG as a storage format.
But, for comparison purposes, everything that is required is that all
the files compared be in the same file format. The information lost in
the compression will of course be even more insignificant in relation to
the differences between the scanners and cameras in question, and they
will be the same for all. To demand TIFF in a comparison is totally
without justification, and this should not be confused with the entirely
rational statement that master files should be recorded and stored
losslessly.
Per Inge Oestmoen, Norway
Bob Monaghan
just a few responses:
lots of people have slides, taken on tripods with good lenses, where they
can just read fine print on signs etc. with high power loupes of 15X or
even 30X. So saying that film doesn't record fine detail doesn't make it
so; film does record it, and it can be seen and printed past 10X
enlargement factors (esp. at the 5 lpmm or 300 dpi quality level). If this
were not so, then there would be no need for loupes past 10X, and they go
far beyond that, just as folks can make bigger prints and enlargements.
And when I project a provia 100 slide on the wall to 8 feet square, and I
can read the fine print in a building sign from 150 feet away, it makes it
hard for me to agree that film can't be enlarged or used past 10X, let
alone less (per 35 lpmm limits from scanned film). ;-)
re: rafe's experiment
Scanning a print at 1600 dpi tells us much about what is on the film?
That's about 1600/25.4 ~= 64 d/mm or 32 lpmm equiv. But prints are not
film, and they are optimized for only about 13 lpmm (with only 8 lpmm
needed to show all the detail the eye can see at normal viewing distances
(based on diagonal of the print). So scanning a print is like scanning a
film of the same size, but a film whose resolution is limited to 13 lpmm
or so.
So when rafe tells us the results are the same, you can't tell the
difference between 4000 dpi scans and a 1600 dpi scan of the print, what
does that tell us? I don't think it tells us much about film limits, but
more about the failure of the scanner to capture the info on film or even
on the print?
conclusion:
If you look at a good quality lab print from a good technique (tripod..)
fine resolution film, and compare to a same size image from a digital 300
dpi printer, the analog image should have truer color and higher contrast.
Contrary to David L., I do believe that you can get enlargements past 10X
with provia 100 fine grain slide film. David's value of 35 lpmm limit for
scanned film (or 40 or 50 lpmm) limits us to a 7X enlargement, at 300 dpi
(5 lpmm).
I think the choice is pretty clear here. Do you believe we can only
enlarge provia 100 by 7X and get a quality image, or do you believe we can
enlarge it to 10X or 15X or more? Can you see fine detail e.g., in signs,
at 15X or 30X on your fine grained film slides, or can't you? If you can,
then that is an example of high resolution and/or high contrast data in a
real world slide ;-) I have lots of slides like that, don't you? ;-)
Bob Monaghan
Now hold on, this is really much simpler than all that ;-)
You do have to compare the original files, without losses. Otherwise, you
can't know if the reason film has lost all its high frequency data after
scanning is the file was stored in a compressed and lossy format.
Similarly, you can't toss away half the color bits using a 16 or 24 bit
color depth format, and then say the 48 bit color drum scanner didn't do
any better than the 24 bit scanner. Maybe it was better and a more
faithful color rendition to the original scene, before you chopped off
those bits.
There is no requirement that you work with compressed or truncated bit
depth files; you can work with the originals, or cropped smaller sections
of them if size is a problem, but in the original full data formats.
regards bobm
David J. Littleboy
>
> Now hold on, this is really much simpler than all that ;-)
>
> You do have to compare the original files, without losses. Otherwise, you
> can't know if the reason film has lost all its high frequency data after
> scanning is the file was stored in a compressed and lossy format.
Low-compression jpeg doesn't lose high-frequency detail. (And even if it
did, there isn't any high-frequency detail to lose. (That much we agree
on<g>.))
> Similarly, you can't toss away half the color bits using a 16 or 24 bit
> color depth format, and then say the 48 bit color drum scanner didn't do
> any better than the 24 bit scanner. Maybe it was better and a more
> faithful color rendition to the original scene, before you chopped off
> those bits.
Grain clump/dye cloud noise means that the valid data in the scan is barely
6 bits, if that. Nothing is lost throwing away 8 bits of noise. Besides, you
can't see 1 bit in 8 differences. If you can see the difference, you're
really in trouble.
> There is no requirement that you work with compressed or truncated bit
> depth files; you can work with the originals, or cropped smaller sections
> of them if size is a problem, but in the original full data formats.
And there's no need to waste disk space and bandwidth with data that
represents noise.
Rafe B.
wrote:
>In article <ru7cuvkif6nngg7qj...@4ax.com>,
Ok, ok. I give up. I've abandoned the "one inch rule." Jeez., who'd
have thought it would cause so much trouble....
I've rewritten the section on "Digicam Captures." Let me know if you
think it's clearer now.
Rafe B.
wrote:
>
>Now hold on, this is really much simpler than all that ;-)
>
>You do have to compare the original files, without losses. Otherwise, you
>can't know if the reason film has lost all its high frequency data after
>scanning is the file was stored in a compressed and lossy format.
Two simple counter arguments to this:
1. The "perfect" scans presented at the top of
the page clearly contain far more detail than
any of the scan samples, even though they
are JPGs of the same dimensions.
2. These JPG images -- for their alleged faults --
clearly reveal differences between different
scanners applied to the same piece of film
(eg., Lawrence Smith's Howtek/Nikon comparison.)
>Similarly, you can't toss away half the color bits using a 16 or 24 bit
>color depth format, and then say the 48 bit color drum scanner didn't do
>any better than the 24 bit scanner. Maybe it was better and a more
>faithful color rendition to the original scene, before you chopped off
>those bits.
There no consensus whatsoever that this extra bit depth is real,
observable, or that it has any impact on the final processed
image. And in any case, my study is focused on spatial resolution
and detail, not tonality.
The notion of 24 bit vs 48 bit image processing is hotly debated
in other forums. FWIW, I work exclusively with 24 bit color.
IMO, the only advantage of the extra bit depth is that it allows more
aggressive tonal manipulation of the image before irreversible
data loss occurs. But by the same token -- if the scan was carefully
captured in the first place -- so as not to *require* aggressive
tonal moves afterward -- the extra bit depth is superfluous.
Wayne Fulton
rafe....@verizon.net says...
>I've rewritten the section on "Digicam Captures." Let me know if you
>think it's clearer now.
Thanks Rafe, yes, I think it is exceptionally clear now.
Rafe B.
wrote:
>In article <rpmcuvc0560nnhtmu...@4ax.com>,
>rafe....@verizon.net says...
>
>>I've rewritten the section on "Digicam Captures." Let me know if you
>>think it's clearer now.
>
>Thanks Rafe, yes, I think it is exceptionally clear now.
Thanks for having a look.
Ti
I didn't mean to make trouble, nor to cause you or anyone in these groups
frustration. On the other hand, I think the few changes you made on your page
make it easier to understand although you are explaining less of your reasoning.
Sometimes, less is more.
Rafe B.
>I didn't mean to make trouble, nor to cause you or anyone in these groups
>frustration. On the other hand, I think the few changes you made on your page
>make it easier to understand although you are explaining less of your reasoning.
>Sometimes, less is more.
No problem, Ti. I thought the "rule" was clear and obvious
but obviously you and Wayne didn't see it that way. The
end result is the same.
rafe b.
http://www.terrapinphoto.com
Michael J Davis
Not sure if you got my point, which is there is a wide variety of
specifications of files called TIFF out there.
The majority *can* be read by the programs I mentioned. However, every
now and then I come across one that cannot.
Mike
[The reply-to address is valid for 30 days from this posting]
--
Michael J Davis
<><
Some newsgroup contributors appear to have confused
the meaning of "discussion" with "digression".
<><
Bill Hilton
>> Netscape uses Quicktime to access TIFF on my system
>> I've never had an unreadable TIFF file cross my machine.
>From: Michael J Davis ?.?@trustsof.demon.co.uk
>
>Not sure if you got my point, which is there is a wide variety of
>specifications of files called TIFF out there.
>
>The majority *can* be read by the programs I mentioned. However, every
>now and then I come across one that cannot.
Michael is right, and you can always find *some* program to read a variant tiff
file (if nothing else, the program that wrote it should always work :).
But there are still a lot of programs that will NOT read 16 bit tiff files or
compressed tiff files, to give two obvious examples. Corel Painter 8 comes to
mind right away, and it's supposed to be a high end graphics program. And
Adobe Photoshop now allows you to save layered tiff files ... try running one
of THOSE on Netscape, I doubt you'll be able to open it.
8 bit uncompressed tiffs usually present few problems for any graphics program,
but compressed tiffs or 16 bit tiffs or layered tiffs (or other variants)
aren't universally accepted by many graphics programs.
Bill
Alan Browne
> Alan Browne <alan....@FreeLunchVideotron.ca> observed
>
>>
>> Michael J Davis wrote:
>>
>>> And these programs include Quicktime, MSpaint, MS Media player,
>>> Photoshop, Irfanview, Netscape, etc, etc.
>>
>>
>> Funny ... Netscape uses Quicktime to access TIFF on my system ... and
>> Photoshop DEFINITELY reads and writes TIFF. I've never had an
>> unreadable TIFF file cross my machine.
>
>
> Not sure if you got my point, which is there is a wide variety of
> specifications of files called TIFF out there.
>
> The majority *can* be read by the programs I mentioned. However, every
> now and then I come across one that cannot.
...well you're an unlucky devil then ... never happened to me that I
could not open a TIFF from any source.
Alan
Bart van der Wolf
>
> David L. gets a value of ~35 lpmm max. from provia 100 when scanned with
> very good technique and prosumer gear. As he and rafe and others have
> noted, prosumer film scanners don't seem to be able to get much beyond
50
> lpmm from scanning images.
I've seen this mentioned several times, but I was too busy to participate
in the thread, until now.
I disagree with the observations mentioned, based on personal experience
with an LS4000 and an SE5400. Something like a quantifyable 70-80 lp/mm
can be extracted by scanning (Provia) film, but it requires a good lens on
a tripod mounted or steady handheld exposure to have so much detail in the
first place. Of course this is only in the plane of optimal focus, not in
the out-of-focus depth of field, or non-flat film area.
> If these are the scanners that you use, you
> won't get the same degree of image quality and enlargeability using
> scanned film as you would from direct printing, IMHO. After all, provia
is
> rated at 140 lpmm (1000:1 hi contrast ratio) and 60 lpmm (1.6:1 low
> contrast), roughly 100 lpmm for typical 6:1 contrast ratio in many
images.
> Even 60 lpmm is waay above 35 lpmm, right? ;-) So if all things were
> equal, then you would expect at least double (60 vs. 35 lpmm) and maybe
> triple (100 lpmm at 6:1 vs. 35 lpmm scanner) the enlargement factor from
> using film directly in enlarger...
SNIP
The only important thing is the combined MTF of lens AND film together
(contrast/modulation near the limiting resolution is important). An
optical microscope can show that there's more fine detail than claimed by
some 4000 ppi scan results. This will be the basis for both optical
enlargement (with unknown lens MTF and focus accuracy) and scanning (with
a good chance on optimal focus due to the collimated nature of most
scanner lighting systems).
Bart
David J. Littleboy
"Bart van der Wolf" <bvd...@no.spam> wrote:
>
> I've seen this mentioned several times, but I was too busy to participate
> in the thread, until now.
> I disagree with the observations mentioned, based on personal experience
> with an LS4000 and an SE5400. Something like a quantifyable 70-80 lp/mm
> can be extracted by scanning (Provia) film, but it requires a good lens on
> a tripod mounted or steady handheld exposure to have so much detail in the
> first place.
http://www.terrapinphoto.com/jmdavis/
Hmm. I still disagree. I just don't see it in real scans. 70-80 lp/mm
implies the same sort of image quality Rafe shows in his "perfect scan"
images. That just doesn't happen in real scans.
> The only important thing is the combined MTF of lens AND film together
> (contrast/modulation near the limiting resolution is important).
I suspect that it's the MTF that's the problem here. Someone might look at
the TechPan scans I put up on that site and jump up and down and shout "75
lp/mm 75 lp/mm!!!", but that doesn't make those scans anything other than
useless mush at much over 10x on a print. 75 lp/mm at under 5% MTF just
isn't useful.
> An
> optical microscope can show that there's more fine detail than claimed by
> some 4000 ppi scan results.
I look at just about every one of my frames with a 60-100x microscope, and
I've never seen anything but the same fuzz that the scanner reports. That
includes the Tech Pan. To the best that I can tell, talking about anything
over 35 to 40 lp/mm on film is head-in-the-clounds ivory-tower fantasy.