Introduction - Feature Extraction and Image Processing - page 10

a computer vision viewpoint. For further details on pattern processing in human vision, see
Bruce (1990); for more illusions see Rosenfeld (1982). One text (Kaiser, 1999) is available
on line ( http://www.yorku.ca/eye/thejoy.htm ) which is extremely convenient.
Many of the properties of human vision are hard to include in a computer vision system,
but let us now look at the basic components that are used to make computers see.

1.4 Computer vision systems

Given the progress in computer technology, computer vision hardware is now relatively
inexpensive; a basic computer vision system requires a camera, a camera interface and a
computer. These days, some personal computers offer the capability for a basic vision
system, by including a camera and its interface within the system. There are specialised
systems for vision, offering high performance in more than one aspect. These can be
expensive, as any specialist system is.

1.4.1 Cameras

A camera is the basic sensing element . In simple terms, most cameras rely on the property
of light to cause hole/electron pairs (the charge carriers in electronics) in a conducting
material. When a potential is applied (to attract the charge carriers), this charge can be
sensed as current. By Ohm's law, the voltage across a resistance is proportional to the
current through it, so the current can be turned into a voltage by passing it through a
resistor. The number of hole/electron pairs is proportional to the amount of incident light.
Accordingly, greater charge (and hence greater voltage and current) is caused by an increase
in brightness. In this manner cameras can provide as output, a voltage which is proportional
to the brightness of the points imaged by the camera. Cameras are usually arranged to
supply video according to a specified standard. Most will aim to satisfy the CCIR standard
that exists for closed circuit television systems.
There are three main types of camera: vidicons, charge coupled devices (CCDs) and,
more recently, CMOS cameras (Complementary Metal Oxide Silicon - now the dominant
technology for logic circuit implementation). Vidicons are the older (analogue) technology,
which though cheap (mainly by virtue of longevity in production) are now being replaced
by the newer CCD and CMOS digital technologies. The digital technologies, currently
CCDs, now dominate much of the camera market because they are lightweight and cheap
(with other advantages) and are therefore used in the domestic video market.
Vidicons operate in a manner akin to a television in reverse. The image is formed on a
screen, and then sensed by an electron beam that is scanned across the screen. This produces
an output which is continuous, the output voltage is proportional to the brightness of points
in the scanned line, and is a continuous signal, a voltage which varies continuously with
time. On the other hand, CCDs and CMOS cameras use an array of sensors; these are
regions where charge is collected which is proportional to the light incident on that region.
This is then available in discrete, or sampled , form as opposed to the continuous sensing
of a vidicon. This is similar to human vision with its array of cones and rods, but digital
cameras use a rectangular regularly spaced lattice whereas human vision uses a hexagonal
lattice with irregular spacing.
Two main types of semiconductor pixel sensors are illustrated in Figure 1.9 . In the
passive sensor , the charge generated by incident light is presented to a bus through a pass