[filmscanners] RE: scanner dmax discussion

["Chuck Cole" <cncole@earthlink.net>]

> -----Original Message-----
> From: filmscanners_owner@halftone.co.uk
> [mailto:filmscanners_owner@halftone.co.uk]On Behalf Of Paul D. DeRocco
> Sent: Friday, July 18, 2003 3:54 AM
> To: cncole@earthlink.net
> Subject: [filmscanners] RE: scanner dmax discussion
>
>
> > From: Chuck Cole
> >
> > The scanner is in no way doing a literal sampling of
> > any kind of mathematical frequency function, and does not perform
> > transforms or inverse transforms as part of its image handling.
>
> Of course a scanner does sampling. It does spatial sampling. All the
> familiar mathematical principles that apply to temporal sampling apply
> equally well, albeit in two dimensions, to spatial sampling. Instead of
> frequency being cycles per unit time, it's cycles per unit length, or
> (fractional) cycles per pixel. You can have aliasing, which on coherent
> spatial frequencies produces moire. You need bandlimiting to
> avoid aliasing,
> which is done with an optical filter. The impulse response of the optical
> filter is the shape of the blur it produces. Dither noise can increase
> apparent bit depth, and break up moire. And so on.

This is true in a philosophical and theoretical sense, but I believe we
would have a difficult time pointing to any scanner which has an electronic
circuit doing anything explicit with samples of an identifiable spatial
frequency function.  At scanner level, it's just an ordered set of pixel
samples that have no explicit extra info about the whole frame, except
something like the wholly passive and static format info of a *.TIFF file
that only conveys the ordering scheme.

These are not coherent optical systems, and optical (as well as verbal :-)
coherence is needed to correspond to your comments on filtering and moire.
Even so, those filters would be fixed and unchanging on a frame-to-frame
basis.  That would tell us no more than citing the difference between one
scanner's A/D being 0-5 volts while another was 0-3.3 volts for the same
full-house bit count: the fact of a simple gain difference should be
irrelevant or utterly trivial.

The scanner does simplistic electronic time sampling of the CCD's data
stream, which is only a set of x values, row by row, in fixed format for the
frame.  Any and all the info that this ordered set of pixel values is an
image is impled, not explict anywhere in this pixel stream or in the scanner
box.  The output file presented to the coumputer doesn't "know" or "say" it
is an image: it is just an ordered set of pixel values with at most some
simple pixel-by-pixel corrections.  I doubt that any scanner products under
$2,000 have DSP or other processors capable of performing in reasonable time
the multi-dimensional, floating point, complex-variable transforms needed to
directly deal with the spatial frequency info of the image.

As I mentioned to Austin, we're way off the deep end on this topic and need
a way to get back to clear relevance for scanners and their explicit active
functions.

Regards,

Chuck

>


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