There might be a little confusion arising from the (earlier List) comparison
of the *causes* of Grain-Aliasing being like superimposing two patterns.
While the analogy is a good one, I'm more comfortable with Tony's previous
suposition that scanners in the 1200-2700ppi range often show G-A because
that arrangement is more likely to have a type of phase vibration than a
higher res (my words, not theirs, since I'm not working directly from the
text of either post).
This seems to be to be what happens in some scans I've seen (and I've seen
my share--if Rafe were using a 2700ppi Acer, he'd quickly become a believer
;-)). A grain of color (usually dark) is large enough to be detected in the
film by the CCD, but too small to register as a pixel, *or* it may be large
enough that it overlaps the fields of two or more pixels but doesn't
completely dominate in any of them. The human eye will 'read' such an area
as a continuous tone, even under a strong loupe (although a microscope would
show that the film grain is in fact there). A scanner, OTOH, is limited to a
digital response--it "sees" the grain but cannot 'split' the pixel into any
more than its 3 RGB components. So it makes a "decision" that can be pretty
far off-base in regards to what the color is really is. Call it "grain,"
call it "noise," call it "Charlie" if you want to--but it looks pretty awful
and it's dificult to deal with, globally.
AFAICS, even if the CCD or Pixel patterns were completely random, this same
thing would happen as long as the pixels were larger than the film grain,
but small enough that no adjacent color would have full dominance in its
field. The driver/program has to make 'decisions,' with only a microsecond
to make them in; sometimes it's wrong.
Anyway, that's the way I see it, and I'm stickin' to my story until somebody
comes up with a better one. ;-)
>From: "Rob Geraghty" <firstname.lastname@example.org>
>Subject: Re: filmscanners: Minolta DiMAGE Scan & Dimage 7 camera
>Date: Sun, 1 Jul 2001 02:15:54 +1000
>"rafeb" <email@example.com> wrote:
> > I also don't really believe in film-grain aliasing --
> > film grain is essentially non-periodic, or, more
> > accurately "white noise" -- ie, containing
> > an even distribution of frequency elements from
> > DC to infinity.
>I don't see why that excludes aliasing of the CCD sample pattern with the
>film pattern, especially with the larger elements of the grain. :-7
Get your FREE download of MSN Explorer at http://explorer.msn.com