[filmscanners] Re: Pixels and Prints

["KARL SCHULMEISTERS" <karlsch@verizon.net>]

David,
I think you have pre-judged the issue and are mixing emotional rhetoric with
supportable and reproducible/verifiable results.

your arguement that on a 'pixel level' film scans aren't the same quality as
10D images is a prime example.
a) its not clear what comparison metric you are using
b) what even 'pixel level' comparison itself means

As was pointed out, a 2 pixel camera has brilliant sharpness, and if it is a
Bayer sub-pixel layout, it will have the 'average' color balance of the
scene capture pretty accurately.  But at best it will be an impressionistic
rendition of the scene.

You also say WRT what kind of comparisons being made
< Someone argued that "scanners produce
better quality pixels because they measure all RGB", and I'm pointing out
that this is wrong because scanned pixels are, in fact, worse than digital
camera pixels.
o o o
Bayer images have very close to the right ratio of luminance to color
resolution for viewing by humans. If you print a Bayer image at a high
enough dpi that you are satisfied with the detail, then the color resolution
will be good enough as well, so the "interpolated pixels" cheap shot is just
that, a cheap shot.>

But other than insisting its 'worse than digital' you really haven't
explained how and why it is worse.  The 'dye cloud' of film acts most
similarly to the Foveon sensor - capturing R,G,B information at each
crystal/sub-pixel location simultaneously.  A Bayer pattern does not.  Which
means that inherently the Bayer pattern sensor is creating data in the
absence of existing data.  Irrespective of how good the estimation equation
is, it is still just that, an approximation.  And like any mathematical
model, it is provably susceptible to uncorrectable artifacts.

The limit of the data that can be resolved on high end film, is higher than
all but the most exceptional digital imagers.  The math just proves this.
The question therefore becomes - does a high resolution scan of this
introduce sufficient optical defects that the end result is a lower data
density than what you can capture from a 10D or similar.  I don't think you
have made the case that it is.  Perhaps for a 1DS, but then you are looking
at a capture device that has a resolution capability that is very close to
the finest grained film data theoretic and empirical limits (just under
4000ppi scan rate).

But the claim that a 10D is better than film isn't supported by the math or
by visual inspection.



----- Original Message -----
From: "David J. Littleboy" <davidjl@gol.com>
To: <karlsch@verizon.net>
Sent: Wednesday, October 22, 2003 7:43 PM
Subject: [filmscanners] Re: Pixels and Prints



"Austin Franklin" <austin@darkroom.com> writes:
>>>>>>>>>>>>>
> I think you've misunderstood what I've said. Take a 900 x 900 pixel crop
> from your 5080 dpi scan and print it at 3x3 inches. Take a
> 900x900 crop from
> a 10D image and print it at 3x3 inches. Which looks better?

That depends,
<<<<<<<<<<<<<<

It doesn't depend. I've never seen a scan that was, on the pixel level, even
close in quality to what the 10D produces.

>>>>>>>>>>>>>>>
 and I am curious why you think that is of any value?
<<<<<<<<<<<<<<<

I'm curious too. I'm not the one making comments to the effect "my
scanner produces 210 MP when your digital camera only produces 6MP".

>>>>>>>>>>>>
  If a 300
x 300 crop from a 10D represents 16x more area, why not compare actual area
for area?
<<<<<<<<<<<<<<

Because that's a different question. Someone argued that "scanners produce
better quality pixels because they measure all RGB", and I'm pointing out
that this is wrong because scanned pixels are, in fact, worse than digital
camera pixels.

(On an area for area basis, it seems digital wins, though. Most people
comparing the 1Ds to 35mm find the 1Ds superior, and I suspect that even a
5080 dpi scan of a 15mm by 22.5mm section of film would look a lot worse at
A4 than a 10D image would.)

>>>>>>>>>>>>>
  You're making the arbitrary choice of sensor sizes/metrics here.
The pixel area from one is not necessarily of equal value to the pixel area
from another, and what the equality is, depends on how many pixels there are
for the respective image.

I could downsample my scanner to give me the exact same image area
information as the 10D, and that information would contain complete color
values, not interpolated pixels.
<<<<<<<<<<<<<<<

Bayer images have very close to the right ratio of luminance to color
resolution for viewing by humans. If you print a Bayer image at a high
enough dpi that you are satisfied with the detail, then the color resolution
will be good enough as well, so the "interpolated pixels" cheap shot is just
that, a cheap shot.

If you print a scanned image at a high enough dpi that you are satisfied
with the detail, then the color resolution will be insane overkill, unless
your audience is Foveon equipped robots. Nothing wrong with insane overkill,
it gets the job done. But it doesn't make a difference in the visual
properties of the print.

If you consider the "minimum dpi for acceptable print" to be a measure of
(the inverse of) an imaging technology's pixel quality, that raises the
question of what is the parameter that limits that minimum dpi. It may be
that it's chrominance resolution that limits dSLR images and luminance
resolution that limits scanned image.

>>>>>>>>>>>
> So the argument that scanned pixels are, on an individual basis,
> in any way
> better than 10D pixels, strikes me as seriously problematic.

But...the 10D doesn't really have pixels...it has sensors, and those sensors
are in a Bayer pattern.  The scanner has full color pixels, and the output
of the scanner can be made to give you pixels that represent the same image
information.  Now, if you want to compare that (and why not, it's pixels for
pixels, which is your metric, and IMO, a far better metric than the
processed output of a digicam vs the "raw" output of a film scanner), then I
guarantee you my 5080 DPI scanner will give me a FAR better looking image
than the 10D will.
<<<<<<<<<<<<<<

Again, I'm not the one comparing pixel-for-pixel; I'm _objecting_ to
pixel-for-pixel comparisons, pointing out that it's a dizzy comparison, and
arguing that you have to downsample scanned images to get comparable pixels
as measured by equivalent print quality.

My best estimate is that 4000 dpi scans of Fuji 100F films downsampled to
2400 dpi turn into close to 10D quality. Maybe the better 35mm lenses are
sharper than my Mamiya MF lenses, and some 35mm scans can be downsampled to
2700 dpi. Whatever. There are lots of people who come up with 9MP or so as
the "digital equivalent" of 35.

David J. Littleboy
davidjl@gol.com
Tokyo, Japan

----------------------------------------------------------------------------
------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title
or body

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body