| dc.description.abstract | Over the past several years, image transmissions and other multimedia
services have become increasingly popular. The problem is that many of
these services are performed in so-called unreliable transmission systems,
for example wireless or mobile systems, where there is a greater
probability of packet loss, bit errors and/or interference when compared to
wired transmission systems.
In today’s transmission systems, retransmission is a commonly used
technique for the correction of lost, or erroneous, data packets. However,
this technique has several shortcomings since the retransmissions often
lead to a considerable amount of extra network traffic, which in turn
results in a reduced availability of bandwidth for each user, and, at worst,
may result in network congestion. The time spent on data retransmissions
may also be a problem in several applications.
This thesis evaluates MD coding as an alternative solution to the abovementioned
problems. MD coding is a technique to achieve robust
communication over unreliable channels such as a lossy packet network. A
source is split into two or more equally important descriptions in such a
way that various reconstruction qualities are obtained from different
subsets of the descriptions. In MD coding, statistical redundancy is added
to a data signal in such a way that data packets, which are lost or exposed
to errors during transmission, can be estimated from all, or some of, the
successfully received data packets. This evaluation is accomplished by
development and implementation of several SD and MD coding
algorithms, where the main objective is to prove that the MD coding
algorithms may considerably reduce the need for retransmission in
unreliable transmission systems. Since the available bandwidth in such
systems is often limited, compression of image data is also an important
part of this thesis.
The developed coding algorithms and compression procedures are
implemented through a test application, which simulates image network
transmissions with user-defined parameters. Finally, the efficiency of each
developed MD coding algorithm is proved by comparing them with the
SD coding systems, the so-called baseline systems.
Through this thesis project, it has been discovered, by means of qualitative
and quantitative tests, that the MD coding systems outperform the SD
coding systems when packet loss occur during image transmission, while
matching the SD coding systems when all the image data are successfully
transmitted. This is done by adding a certain amount of redundancy to the
MD data signals. | en |
