Image Processing Reference
In-Depth Information
the horizontal transport registers. This is the column bus supplied by the pixel sensors. The
horizontal transport registers empty the information row by row (point by point) into a
signal conditioning unit which transforms the sensed charge into a voltage which is
proportional to the charge in a bucket, and hence proportional to the brightness of the
corresponding point in the scene imaged by the camera. CMOS cameras are like a form of
memory: the charge incident on a particular site in a two-dimensional lattice is proportional
to the brightness at a point. The charge is then read like computer memory. (In fact, a
computer memory RAM chip can act as a rudimentary form of camera when the circuit -
the one buried in the chip - is exposed to light.)
There are many more varieties of vidicon (Chalnicon etc.) than there are of CCD
technology (Charge Injection Device etc.), perhaps due to the greater age of basic vidicon
technology. Vidicons were cheap but had a number of intrinsic performance problems. The
scanning process essentially relied on 'moving parts'. As such, the camera performance
changed with time, as parts
wore
; this is known as
ageing
. Also, it is possible to
burn
an
image into the scanned screen by using high incident light levels; vidicons also suffered
lag
that is a delay in response to moving objects in a scene. On the other hand, the digital
technologies are dependent on the physical arrangement of charge sites and as such do not
suffer from ageing, but can suffer from irregularity in the charge sites' (silicon) material.
The underlying technology also makes CCD and CMOS cameras less sensitive to lag and
burn, but the signals associated with the CCD transport registers can give rise to
readout
effects
. CCDs actually only came to dominate camera technology when technological
difficulty associated with
quantum efficiency
(the magnitude of response to incident light)
for the shorter, blue, wavelengths was solved. One of the major problems in CCD cameras
is
blooming
, where bright (incident) light causes a bright spot to grow and disperse in the
image (this used to happen in the analogue technologies too). This happens much less in
CMOS cameras because the charge sites can be much better defined and reading their data
is equivalent to reading memory sites as opposed to shuffling charge between sites. Also,
CMOS cameras have now overcome the problem of
fixed pattern noise
that plagued earlier
MOS cameras. CMOS cameras are actually much more recent than CCDs. This begs a
question as to which is best: CMOS or CCD? Given that they will both be subject to much
continued development though CMOS is a cheaper technology and because it lends itself
directly to intelligent cameras with on-board processing. This is mainly because the feature
size of points (pixels) in a CCD sensor is limited to about 4 µ m so that enough light is
collected. In contrast, the feature size in CMOS technology is considerably smaller, currently
at around 0.1 µ m. Accordingly, it is now possible to integrate signal processing within the
camera chip and thus it is perhaps possible that CMOS cameras will eventually replace
CCD technologies for many applications. However, the more modern CCDs also have
on-board circuitry, and their process technology is more mature, so the debate will
continue!
Finally, there are specialist cameras, which include
high-resolution
devices (which can
give pictures with a great number of points),
low-light level
cameras which can operate in
very dark conditions (this is where vidicon technology is still found) and
infrared
cameras
which sense heat to provide thermal images. For more detail concerning camera practicalities
and imaging systems see, for example, Awcock and Thomas (1995) or Davies (1994). For
practical minutiae on cameras, and on video in general,
Lenk's Video Handbook
(Lenk,
1991) has a wealth of detail. For more detail on sensor development, particularly CMOS,
the article (Fossum, 1997) is well worth a look.
