WD said:
So you are saying in fact that with an incoherent source the MTF
cutoff
actually is at a higher spatial frequency than with a coherent source,
but in the 'region of interest'
(spatial frequencies above Nyquist cutoff of CCD array)
there is an attenuation of spatial frequencies.
Yes. In the "region of interest", as you put it, the MTF of the
semi-coherent/collimated source is much higher than the diffuse source.
Another related question I have regards scattering effects.
For example the Callier effect, small trapped air bubbles
(peppergrain),
etc. Your above discussion made no differentiation regarding
scattering
vs. pure transmission/absorption. For example if there was an element
(say grain or tiny bubble), which corresponded to spatial frequencies
beyond the Nyquist freq. of the CCD array, your above explantion
compares
the effect of collimated vs. non-collimated light sources whether or
not
these elements absorb or scatter the light.
That is correct - the previous discussion centred on the effect of MTF
as a function of the spatial frequency.
How does this play into
the picture? Is it merely a matter of the resultant amplitude of the
nasty
spatial frequencies in question? (scattering element vs.
non-scattering element)
The Callier Effect has an impact on contrast that is essentially
constant across the spatial frequency band. Scattering occurs from
grain centres independently of whether the grain centres are part of a
large object or a small high resolution part of the image.
On a negative, the CE has the effect of increasing contrast - the dense
parts of the negative scatter more, and thus appear more dense, whilst
the light parts of the negative scatter less and thus appear lighter. In
a slide the effect is reversed and, of course, the effect is almost
negligible with diffuse sources, since light that is scattered out of
the field of the projection lens is almost perfectly balance by light
that would normally be travelling out of the field but is scattered into
it.
However, the CE is virtually negligible with colour emulsions and this
is fairly obvious by scanning a grey ramp on a slide (or colour
negative) in a collimated source scanner and then observing the colour
balance throughout the ramp.
Scattering from particles such as grain is a strong function of the
wavelength of the light and the size and transparency of the particle.
One of the commonest examples of this is the clear blue sky and the red
sunset that everyone is familiar with. The sky is blue because the
short blue wavelengths of light scatter more easily from particles in
the atmosphere than the longer red wavelengths. Thus, in the clear sky
the only light that reaches your eye is sunlight that has been scattered
by the atmosphere. Without that scattering, the blue sky would be black
- and indeed, because the scattering also produces polarisation, it can
be made to appear a much darker blue with a crossed polariser.
Similarly, the setting sun appears red because the shorter blue
wavelengths have been scattered out of the direct line of sight to the
sun, leaving only the red. As the density and size of the scattering
centres increases, , even the longer red wavelengths become scattered,
so the cloudy sky appears white, grey or even black, and all of the
direct rays to the setting sun become scattered and it cannot be seen.
Thus, with the grey ramp on the film one would expect to see a strong
change of colour as the ramp increases in density - with it initially
becoming yellow then redder (or more cyan, if a negative) and then
returning to neutral balance, with reduced density, finally the red rays
are scattered. This doesn't appear, at least in any tests I have made,
to be the significant colour change, and is much less than the general
bias of the film itself. Note that this colour shift should be
independent of the film type, yet we all know that the choice of film
can dictate the general balance of shadows and highlights much more than
any general trend that the CE would suggest should occur.
This is all very interesting. I have been playing around with a
diffuser
with a coolscan 5000 and have found in certain cases dramatic
improvements.
What surprises me is that scanner vendors are so slow to catch on.
Nikon seems not yet to have caught on, Minolta presumably has with the
5400,
Imacon in their latest and greatest top of the line scanner is touting
its diffuse light source.
I would guess that the increase in scan times from a marketing
perspective
explains part of it,
That is true, and is more significant for Nikon than other manufacturers
because of the light source they use. Nikon have already optimised
their scanners for speed, for example choosing a low f/# optic with
limited depth of field in preference to something that would give a
longer scan time.
but with such potential improvement (in certain
cases),
it still surprises me.
Me too! It isn't as if this is new to scanning - although the effect of
aliasing the grain may be - the principle has been understood for the
best part of a century or more.
