Mike said:
As you approach the Nyquist frequency, certain frequencies are reduced in a
very predictable way. This is softness.
This is a contradiction of your earlier statement, since scanning at a
higher resolution (ie. increased sampling density) results in a higher
Nyquist limit and thus, by your latter argument, shifts the onset of
this "softness" to higher spatial frequencies in the image. In short,
your latter argument indicates that scanning at a higher resolution
results in *less* softness, not more, in an image scaled at the same
size! Which argument are you making?
The reproduction of spatial frequencies are reduced by the MTF of the
scanner, which decreases not only as you approach Nyquist, but
throughout the spatial frequency range, usually monotonically from a
maximum at zero cy/mm. Sampling density merely determines where on
that MTF curve the Nyquist limit sits. As sampling density increases,
more of the total MTF range is included in the spatial frequency range
that can be unambiguously reproduced - so more information is resolved,
not less. Clearly this is an issue of diminishing returns, but the
result is always positive - more total detail resolved, not less and
certainly not more "softness". In some systems it is possible to sample
such that the Nyquist limit lies beyond the limiting MTF of the scanner
(eg. most flatbed scanners) thus meeting the criteria for total
elimination of aliasing. In such cases, increasing the sampling density
will not gain resolution, but neither will it increase image softness -
the end result is just more data representing the same image information
content.