[filmscanners] Re: Nikon Coolscan

["Anthony Atkielski" <anthony@atkielski.com>]

Arthur writes:

> If I buy a camera and I thought it was supposed
> to come with a chocolate Easter bunny, and I ask
> someone, "was I supposed to get a chocolate
> Easter bunny with my camera?"  The response, "Why
> do you need a chocolate Easter bunny, it won't
> do anything to help you take better pictures"
> is not answering his question.

But it is extremely relevant to the general topic of photography or cameras,
whereas the initial question about a chocolate Easter bunny is not.  So the
reply actually makes more sense than the question.

> Are you saying that a 4000 ppi scan of a Provia
> 35mm frame (making a nearly 6000 x 4000 pixel scan)
> is equivalent to a digital camera image that is
> 3000 x 2000 pixel using the Bayer matrix, meaning
> it is really a much lower resolution which is
> interpolated?

In terms of resolution, a 6000x4000 film scan does not actually contain
6000x4000 clearly distinct pixel values.  The contrast between adjacent
pixels is quite low in most cases.  So, while details only a pixel wide are
often picked up, they differ in contrast only slightly from their neighbors,
which lowers useful sharpness overall.  When the image is downsampled, the
result is very sharp, since contrast between adjacent pixels is much higher.
I don't think it need be downsampled to half the size, though.

As a test, I've compared direct scans at 4000 dpi with digital images that
have been _upsampled_ to the equivalent resolution.  The direct scans of
film win every time, especially with respect to color resolution (thanks to
the wicked matrix filter in single-CCD digicams), but also with respect to
luminance resolution.  So, while the original scans do look blurry up close,
they still contain more information than a digicam image does.  But the way
to see this is not to downsample the film scan (which simply gives you two
images with less information) but to upsample the digicam image (which shows
the lack of information in the digicam image).

> Are you therefore suggesting that a 4000 ppi scan
> of a 35mm frame of Provia has captured all the film
> has to offer ...

No.  It's just that it is very difficult to scan at 4000 dpi and guarantee
that each pixel recovered will have infinite contrast as compared with its
neighbors.  It's a scanner limitation.  Still, _most_ of the detail in
Provia is captured at 4000 ppi; in fact, unless the shot was taken on a
tripod with a very good lens, the scan may well contain everything there is
to recover from the slide, in which case scanning at higher resolutions
gains nothing.  For films with lower resolution, or handheld shots, or
cheaper lenses, 4000 dpi is plenty; for ISO 800 film, I get just about
everything there is at 2700 dpi.

> Or that 4000 ppi scanners are just not capable
> of capturing film information at better than
> the equivalent of a 3000 x 2000 pixel bayer
> interpolation?

They do better, but not as much better as the difference in resolution
implies, because the contrast between adjacent pixels is limited.

> What do you believe is the maximum resolution
> of the information on the Provia 35mm film frame?

For high contrast images (1000:1, pretty much unknown in the real world),
you may be able to resolve additional detail up to 6000 dpi or so.  For more
ordinary contrast, 4000 dpi is more than sufficient.  In most ordinary
photographs, very few details show enough contrast to resolve anything at
more than 4000 dpi, and probably 95% of the detail can be resolved at 2700
dpi.

> Do you think the limiting factor in scanning
> film is the scanner's ability to capture more res,
> or that the film doesn't offer anymore to be
> captured due to either the film or the res
> of the 35mm camera lenses, or film flatness or
> hand movement or other mechanics or optics that
> reduce final image resolution?

At less than 6000 dpi, and under _ideal_ conditions, the scanner is a
limitation.  In real-world practical photography, however, very few images
can justify a 6000-dpi scan; 4000 dpi is plenty, and 2700 dpi is good enough
for many purposes.

This assumes, though, that the scanner is really pulling high contrast in
adjacent pixels.  If the scanner isn't good at this, and contrast is quite
low, you might have to scan at higher resolution in order to get the
necessary contrast at lower resolutions.  In other words, a cheap scanner at
4000 dpi might be no better than a very good scanner at 2700 dpi.  This, I
think, is why some high-end scanners with seemingly lower resolutions still
manage to produce nicer, sharper scans.  Not much point in having a nominal
resolution of 6000 dpi if, say, the optics in the scanner can't resolve more
than 4000 dpi.

> Do you think that there is therefore no point
> in improving resolution on, not only film scanners,
> or that scanner technology is simply not able to
> currently capture anything approaching the resolution
> specs being claimed?

Resolution is not a yes/no concept.  Most scanners do not capture at their
nominal maximum optical resolutions with very high contrast, but they do
resolve details at those maximum resolutions.  In some cases, the details
resolved are not quite as fine as those on the film, and do not show quite
as much contrast as they do on the film.  Again, however, this assumes ideal
conditions for the film.

> ... do digital cameras which can capture above
> the 2000 x 3000 bayer matrix (equivalent to 35mm
> in dimensions) image size actually surpass 35mm
> film frames in resolution?

For luminance only, not unless they are doing 6000x4000 or beyond.  For
color, not unless they are doing about 12000x8000 or beyond (because the
matrix filter greatly reduces color resolution as compared to film, and so
you need a great many pixels to compensate for this).



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