Natural Science Forum / Physics / Optics / October 2005

It's important to understand that it is possible to improve image
quality without somehow increasing the 'amount of information'. To put
it another way: filtered images can be more useful or simply 'look
better'.  The elimination of jagged edges is an example of this.

Here's another example: Suppose you'd like to know the precise location
of some feature in an image, say, the center of small round feature or
the edge of something. (These problems are commonly encountered in
astronomy and machine vision, respectively.) Unless you use some
technique like interpolation, your resolution will be limited to pixel
pitch. However, the resolution improvement afforded by simple
interpolation is often limited by noise in the image data (not to
mention the relationship between the pattern of light from the source
versus the interpolation model).. Feature location estimation
algorithms that transform data from larger numbers of pixels have the
potential of providing improvements in location estimation proportional
to the square root of the number of pixels involved (assuming random
noise and central limit theorem applicability). Astronomers actually
defocus their telescopes at times, in order to spread the image of a
star over a few dozen pixels. The resulting 'blurry' image is the basis
for more accurately estimating the centroid location. (With the
availability of multi-megapixel imaging chips, I don't know if this
technique is still used.)

Returning to the discussion of scanners: Most use line arrays in one
dimension and stepper motor drive in the other. For the motorized
dimension, it is possible to vary the effective pitch by microstepping,
but I suspect that most control schemes simply vary the speed and
synchronization relative to the clocking rate for the other axis. As
another poster pointed out, the USB 1.1 and parallel port interfaces
used on many scanners can be a bottleneck that has variable effects on
throughput and the start/stop behavior of the scanning mechanism. Most
scanners rely heavily on resources within the PC.
Paul Mathews