Page 240 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 9.6. Forward Techniques 217

the bitstream, the decoder discards all bits until the next resynchronization
codeword, where synchronization is reestablished and the decoder resumes its
decoding process. The discarded bits may well be correctly received but cannot
be decoded correctly due to loss of synchronization. In the case of RVLCs,
when the decoder identi es the next resynchronization codeword, instead of
discarding all preceding bits, the decoder starts decoding in the reverse di-
rection to recover and utilize some of those bits. This is illustrated in Figure
9.3(b).
Reversible variable-length coding has been adopted in most recent stan-
dardization e orts. For example, the modi ed unrestricted motion vector mode
(modi ed annex D) of H.263+ uses RVLC to encode motion vector di er-
ences, the data partitioned slice mode (annex V) of H.263++ uses RVLC to
encode header and motion information, and MPEG-4 uses RVLC to encode
texture information.

9.6.4 Layered Coding with Prioritization
In layered coding, video is encoded into a base layer and one or more enhance-
ment layers. The base layer is separately decodable and provides a basic level
of perceived quality. The enhancement layers can be decoded to incrementally
improve this quality.
Layered coding can be useful when applied over heterogenous networks
with varying bandwidth capacity. However, to be used as an error-resilience
tool, layered coding must be combined with prioritized transmission or what
is commonly known as unequal error protection. In this case, the base layer
is transmitted with higher priority or a higher degree of error protection. For
example, in Ref. 186 Ghanbari introduced the concept of layered coding with
prioritized transmission to increase the robustness of video against cell loss
in ATM networks. In this technique, the encoder generates two bitstreams.
The base-layer bitstream contains the most vital video information, whereas
the enhancement-layer bitstream contains residual information to improve the
quality of the base layer. The base layer is then transmitted using high-priority
ATM cells, whereas the enhancement layer is transmitted using low-priority
cells. When tra$c congestion occurs, low-priority cells are discarded rst.
Another example is the power control method proposed in Ref. 187. In this
method, when video is transmitted over a wireless network, more power is
used to transmit the base layer, whereas less power is used to transmit the
enhancement layers.
There are many ways to encode video into more than one layer. For exam-
ple, the base layer can include a low-frame-rate version of video, whereas the
enhancement layers can contain frames used to increase the frame rate. This is
usually referred to as temporal scalability. Another method is when the base