At 5:52 AM +0000 9/23/02, olympus-digest wrote:
>Date: Sun, 22 Sep 2002 21:48:35 -0700
>From: dreammoose <dreammoose@xxxxxxxxx>
>Subject: Re: [OM] Grain aliasing
>
>Joe Gwinn wrote:
>
> ><> The question that nagged me while I was reading was simple: Given that
> >film grain is totally random, why would scanning (sampling at some number of
> >pixels per inch) cause grain to be affected more than anything else in the
> >image? Random is random, so the scanner pitch (number of pixels per inch)
> >shouldn't matter one bit. And the grain is far too fine for even the pro
> >scanners to resolve. Yet, it seems to matter. Why?
> >
>What size is film grain? Some of you folks know!
>
> ><> Film granularity (grain clumping)
> >
>And what size is the granularity?
According to "Photography - Its Materials and Processes", sixth edition,
C.B.Neblette, Van Nostrand 1962, Chapter 24 "The Microstructure of the
Developed Image":
One can resolve the individual silver grains at 2500x magnification, and using
an electron microscope at 25,000x, one can see the filamentary structure of
those grains. These are far beyond the magnifications used in photographic
enlargement. The film tests in the magazines talk of perhaps 40x, so the
"grain" one sees in these test shots are in fact granules.
The granules are far larger, but harder to characterize simply. As a
densitometer aperture is scanned slowly across a piece of film developed to
constant average density, the actual observed density will vary randomly due to
the granularity (not the actual silver grains, despite the usual practice of
calling this the "grain"). With a sufficient number of density samples, one
can compute the mean (average density) and the standard deviation about that
mean. The standard deviation is a measure of the width (spread) of the
distribution of densities. This is also known as the root mean square (rms)
variation.
The basic theory is that the product of the standard deviation of density and
the diameter of the scanning aperture used to measure density should be a
constant regardless of the diameter of the aperture; this constant is called
the granularity "G" and is characteristic of the film and developer used. (It
also varies with the average density. For instance, overexposure seems to
cause excess grain in many film types.)
>From the Kodachrome 64 datasheet: Using a aperture 48 microns (0.0019") in
>diameter, equivalent to 12x magnification, with the test film developed to
>have a density of 1.0 (10% transmission), the diffuse rms granularity is "12".
> What the datasheet does not say is what the scaling and units of 12 are, but
>there is a Kodak technical note explaining the number:
><http://www.kodak.com/global/en/professional/support/techPubs/e58/e58.shtml>.
The number 12 is 1000 times the rms (root mean square) variation in the
measured density, so what they are saying is that the average density is 1.000
and the standard deviation around that is 0.012 in density units (the base ten
logarithm of the transmittance), for a 48-micron aperture.
So, translating into a language somewhat closer to English, the average
transmittance is 0.10 (that is, 10%), and 680f the density samples fall in the
range 0.1028 to 0.0973 (10.3% to 9.73%) transmittance.
Q=(48)(12)= 576
However, one cannot deduce a typical granule size from this, although the
smaller the aperture (the larger the equivalent magnification), the larger the
effect of granularity (larger the rms value).
To get the typical granule size, one wants the autocorrelation function of the
densitometer scan, so I did a Google search, finding a very technical report on
scanners used for photogrammetry:
<http://phot.epfl.ch/workshop/wks96/art_1_2.html>. This report shows the
diffuse light sources work far better than directed (kohler) light sources,
confirming a widely held suspicion.
Finding the autocorrelation data will take a while, it seems. My first
attempts were not successful. Does anybody have scans of a blank wall showing
bad granularity? A TIFF or BMP file could be analyzed. (JPEG and GIF are
compressed, screqing up the analysis.)
> ><> Now, we can see where the grain aliasing comes from. First of all, it's
> >really granularity aliasing; the grains are too small for any non-research
> >scanner to see. And (unlike the grain) the granularity has a built-in
> >characteristic distance, the average spacing between granules, and if the
> >scanner's sampling pitch is about the same as the average spacing between
> >clumps, the effect of the granularity will be greatly enhanced, and may
> >become wierd as well.
> >
>Is wierd weider than weird?
The spelling of weird is weird.
> >What to do? Films designed to be scanned will have somehow abolished the
> >peak, so there is no characteristic average distance, or moved the
> >characteristic distance well away from typical scanner pitches, probably by
> >making the characteristic distance far smaller the the scanner pitch.
> >
>So a film 'optimized' for scanning would be optimized for scanning at a
>particular optical dpi.
Or a set of dpi values found in widely used scanner brands. There probably are
not that many distinct values.
>This might explain the recent post about how
>scanning optimized film performed worse for one listee's particular
>scanner than another 'normal' film. It might also explain the
>differences in results between listees with the same films, different
>scanners. Any theory on the likely effects of coherence of the scanner
>light on this phenomenon?
The Callier Effect is independent of the grain aliasing effect, except that
(according to Kodak) kohler illumination systems will cause the roughness of
the film base surface to be seen as added granularity.
>Would changing the 'clump size' affect edge characteristics, and thus
>apparent sharpness? When people talk about finer grained films with less
>apparent sharpness, are they talking grain or granularity?
Granularity only; true grain is far too small to be seen, so edge effects et al
are all effects on the clumps.
> >If one has existing film to be scanned, the only solution is to change the
> >scanner pitch (optical, not interpolated) to avoid the peak.
> >
>So it is possible that a relatively lower dpi scanner, possbily with a
>low coherence light source, would likely have success with a higher
>percentage of random existing film images?
That's the implication, and the hope.
What would also work would be slight defocus of the scanner, such as that
caused by imposition of a sheet of slightly diffuse plastic between photo and
scanner.
> >If my theory is correct, making the scanner optical pitch larger or smaller
> >will work equally well, though one's instinct is to go for finer.
>
>It will all depend on the actual probability density versus distance function,
>which may not have just one peak. Cheaper to buy a couple of scanner, maybe
>2700 and 4000dpi than a non-existant variable optical pitch scanner?
I don't know that my theory is correct, rather than merely plausible. One test
would be to scan a set of sample photos taken on various kinds of film on
various scanners, and compare the results.
Joe Gwinn
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|