Algorithms of MD - Distributed Multiple Description Coding - page 91

Fig. 4.33 Block diagram of the proposed scheme

tion between frames, the redundancy is added adaptively to make a better trade-off
between the reconstructed quality and the compression efficiency. Consequently, the
abrupt motion may result in worse side distortion so more redundancy is needed to
guarantee the correlation between frames, while less or no redundancy is needed if
the motion of inter-frame is smooth enough.

4.5.2 Proposed Scheme

Figure 4.33 illustrates our proposed scheme for MD video coding. In the preprocess-
ing stage, the original video sequence is up-sampled to generate a new-length video
with adaptively redundant frames. And then by means of odd and even frames, the
new video sequence is divided into two descriptions, which can be compressed by
any standard video codec. Here, the latest video coding standard H.264 is employed
for a suitable comparison. In the post-processing, the decoded video stream is down-
sampled to reconstruct original-length video, and error concealment method is used
to estimate lost frames. The details of pre- and post-processing are showed in the
following subsections.

4.5.2.1 The Preprocessing Stage

In terms of the principle of MDC, higher quality of side decoded video that
will result from more correlations in descriptions for better error concealment
is available, but more redundancy introduced will bring about lower efficiency to
the central decoder. Obviously, it is a better solution that the redundancy added can
make a trade-off between the reconstructed quality and the compression efficiency.
As a result, in the preprocessing stage, the up-sampling with changeable rate is
employed to introduce the adaptive redundancy. Since different motion appearance