act
ic
ite
vay
Lng
ic
d,
ind
ng
Ire
"ms
en
it
Ch
be
in
re
st
.nt
th
h-
ve
al
th
an
ire
re
ime
‘he
is
il.
)l-
‚th
In order to answer these questions we need a model for the digitizing
process that allows us to make an analysis and decide what parameters
are of interest. Here, firstly quality measures of photographic images
and their relation to the digital image will be considered. Then a
general model of a digitizer is presented. The model consists of two
parts. The first models the quantization in space and the second
quantization of the grey levels in the image.
The models are based on statistical optics and signal theory. Some of
the fundamental theorems are valid only when restrictions are imposed
on the input signal or the image. Working with real images these
underlying assumptions will sometimes be violated.
3. RESOLUTION AND PIXEL SIZE
3.1 Sampling in the Space Domain
The digital image is constructed from the photograph by sampling at
regular intervals. In each sampling point the intensity of the image
is measured and converted to a number. According to the sampling
theorem (O'Neill, 1963) it is sufficient to sample a function with an
interval A < 1 / 2R, where R is the maximum frequency. Such sampling
permits complete reconstruction of the original function and no
information is lost.
When digitizing an image the interval A equals the distance between
the pixel centers. The formulation of the theorem quoted above applies
to onedimensional signals. The image considered as a signal is
two-dimensional. The frequency associated with images is the spatial
frequency or lp/mm. With square pixels and arbitrary orientation of
the lines in the image, the sampling must be made with the interval:
A« 1:2 SER. ( 3.1 )
The sampling theorem is valid under the condition that the input
signal is band limited. The lens in an optical imaging system works as
a low-pass filter in the frequency domain. As a result the image will
always be band limited. The problem is how to estimate R in the for-
mula above to permit a proper choice of the sampling interval A.
3.2 Resolution. Modulation Transfer and Threshold Modulation
Image quality and performance of photographic systems are evaluated
in terms of resolution, Modulation Transfer Functions (MTF) and
granularity ( Brock, 1970 ). Interpreting frequency as lp/mm we can
use resolution to decide how to sample the image in accordance with
the sampling theorem quoted in the preceding section. Then the choice
of R, or cut off frequency will give the appropriate sampling
interval, or pixel size A. Working with lenses for photogrammetric
purposes, in most cases the lens is the limiting factor. Thus the
resolution of the lens could be taken as the criterion. A still better
solution is to use system resolution.
The Modulation Transfer Function describes how modulation transfer in
141