Information Technology Reference
In-Depth Information
Data Averaging or Sampling: Rather than store information for
each pixel in a picture, you can condense the size of an image file by
selecting a smaller collection of information to store. As a simple
(but not very good) example, you could choose to store only every
other pixel in an image. Because only half the pixel information
would be stored, the size of the image file would be cut in half from
the original. However, when you decide to display the full picture,
your computer would be forced to fill in the missing data. One pos
sibility your computer may take would be to assume that the color
at one pixel was just the same as the one just before it. Effectively
this would display each stored pixel over two parts of the reconsti
tuted image. Your computer might instead decide to compute the
average color numbers of neighboring stored pixels to determine the
value to be displayed at a nonstored position. In general, selection
of only certain pixels to store is called
sampling
. Reconstruction of
pictures by averaging is called
data averaging
.
With any such sampling or averaging approach, the resulting
picture cannot be as sharp as the original, because some informa
tion has been lost. The resulting picture still might be recognizable;
it might even seem quite acceptable for many purposes. However, it
will have lost some crispness and clarity from the original. As aver
aging or sampling increases, picture quality can be expected to dete
riorate further.
Voice Sampling
Although this notion of sampling may seem like an impractical and unrealistic option for
condensing space, it should be noted that this process is common in another context:
voice communications over the telephone. To be more precise, when you speak, you
create sound waves that are picked up by the telephone's microphone; when you listen,
you hear sound waves created by a speaker. These are analog signals, and ordinary
telephones may communicate directly with analog signals. At some stage in communica
tions, however, these analog signals may be converted to a digital format and converted
back later. In this digital format, sampling can be used extensively.
To analyze our talking, we could consider our speech as producing an infinite series of
values concerning pitch and volume. Rather than transmitting this continuous stream of
speech, telephones typically sample speech about 8000 times per second, using 8 bits
of information for each sample. This produces 64,000 bits of information per second,
rather than the infinite stream of communications from the person talking. Although the
details are technically complex, the bottom line is that telephone systems can recon
struct voice patterns quite well by sampling, rather than transmitting the continuous
(and infinite) speech patterns of a speaker.
