[filmscanners] RE: over resolving scans

["Joe Pedit" <pedit@email.unc.edu>]

I've attempted to characterize grain (grain aliasing?) for my Sprintscan 4000 
and a few films
including Provia and Tech Pan.  I generated semivariograms for the red, blue 
and green channels
of Tech Pan (9 minutes in D76), Provia 100F, Supra 400, and Royal Gold 1000.  
The results for
each channel are at:

<http://www.unc.edu/~pedit/red.gif>
<http://www.unc.edu/~pedit/blue.gif>
<http://www.unc.edu/~pedit/green.gif>

The y axis is the semivariance and the x axis is the lag measured in pixels.  
In general, the
sills increase as expected (i.e., Provia 100F < Tech Pan < Supra 400 < Royal 
Gold 1000).  It is
difficult to analyze the spatial component of grain (grain aliasing?) in the 
semivariograms due
to the vastly different sills.  I normalized the data by the sills so the 
spatial trends are
more easily seen.  The results are at:

<http://www.unc.edu/~pedit/red_normalized.gif>
<http://www.unc.edu/~pedit/blue_normalized.gif>
<http://www.unc.edu/~pedit/green_normalized.gif>

In general, Provia 100F needs the largest lag (3 to 4 pixels) to reach its 
terminal sill value
even though it has the largest normalized semivariance at a lag of one pixel.  
The negative
films reach their sills within 2 to 3 pixels.

It would be interesting to repeat this analysis with scans from lower and 
higher resolution
scanners to quantify the effect of scanner resolution on aliasing.

Information on semivariograms is at:
<http://w3eos.whoi.edu/12.747/notes/lect05/l05s03.html#variograms>

Joe Pedit

> Topic: [filmscanners] RE: over resolving scans
> =========================================
> Date: Tue, 8 Oct 2002 23:04:54 +0200
> From: "Anthony Atkielski" <anthony@atkielski.com>
>
> At certain key scanning resolutions, which vary by film, exposure, etc.,
> there is some interaction between the scan and the grain in the image that
> may create the impression of larger or more obvious grain in the scan than
> is actually present in the film.  If the resolution is held below this key
> threshold, grain simply blurs out of existence (but detail does, too); if
> the resolution is raised above this threshold, the interaction disappears
> and grain starts to look as it would on an optical enlargement.
>
> This seems to happen a little bit with Tri-X at 2700 dpi, although it's
> really hard to tell.  It is rumored that this happens with Provia and Velvia
> at much higher resolutions, which may be one reason why they seem to show
> more grain in scans than they actually have in real life, but I've not been
> able to isolate clear evidence of this myself.
>
>
>> He may prefer the mushier grain appearace to
>> the sharper more emphasized one; whereas others
>> do not.
>
>
> I actually prefer no grain at all, but few films have such fine grain that
> it cannot be detected even in the highest-resolution scans.  Tech Pan is in
> this category.  Also, interestingly, Portra 400BW and some other chromogenic
> B&W films have incredibly fine grain, especially given their high speed;
> 400BW has grain that looks a lot like Tech Pan, at least at 4000 dpi or
> below (that is, it appears to have practically no grain at all).  This has
> persuaded me to shoot a lot of 400BW; I don't like the way it reacts to
> light, but the absence of grain is great, and allows cropping and
> considerable enlargement.  For a good example of how well this film can
> perform under good conditions, see


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