Databases Reference
In-Depth Information
Information Source
Channel
x(t)
x
n
x
c,n
Filtering
Information
Source
coder
Channel
and
coder
cource
sampling
Channel
y(t)
y
n
Source
decoder
Channel
User
Reconstruct
^
c,n
decoder
User
F I GU R E 8 . 1
A simplified block diagram of a communication system.
sources. We will describe several different models that are widely used in the development of
lossy compression algorithms.
Finally, let us briefly look at the context of the compression operation. A very simplified
block diagram of a communication system is shown in Figure
8.1
. An information source
generates a signal (often an analog signal). In the figure we have represented the signal as a
function of time. This would be the case if the signal was a speech or audio signal. However,
the information signal may be a two-dimensional spatial signal such as an image, or a function
of both space and time such as video. If this information is not in discrete form it is discretized
to generate a sequence of values. Again we have depicted this signal as being a function of a
single index as would be the case for a speech or audio signal. However, it could just as well
be a function of two or three indices. This discrete sequence is the input to the compression
algorithm, or in more general terms, the source coder, which attempts to reduce the average
number of bits per sample used to encode the sequence. The process of source coding usually
results in the removal of redundancy in the signal, making it vulnerable to noise. In order to
protect against errors the output of the source coder is encoded using a channel encoder that
introduces redundancy in a controlled fashion. The block labeled “
channel
” is usually thought
of as the medium over which the information is transmitted from one location to another. In
the case of something like a cell phone the locations may be spatially distinct—the information
is transmitted from “here” to “there.” However, in many cases, such as when we store the
information to be examined later, the locations my be temporally distinct—a communication
from “now” to “then.” We will generally only deal with the discrete form of the data and then
only the source coding aspect, so we tend to simplify this diagram even further by lumping
together the blocks in the dashed boxes. Thus we have an information source generating a
sequence of samples
which are encoded by the source coder and transmitted, spatially or
temporally, over the channel. We will assume in this topic that the channel does not introduce
any distortion so the source coder output is identical to the source decoder input. The source
decoder output
{
x
n
}
is then supplied to the user. The goal of the source coder is to use as few bits
as possible to encode the information-bearing sequence
{
y
n
}
{
x
n
}
while still keeping the reconstruc-
tion sequence
{
y
n
}
as a close approximation. We are using very fuzzy language here when we
