Page 233 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

210 Chapter 9. Error-Resilience Video Coding Techniques

sequence number sub eld. This sequence number can be used to detect packet
losses at the channel decoder. Similarly, the group-number (GN) codeword in
an H.263 GOB header can be used to detect errors at the source decoder.
Another method that can be used by both the channel decoder and the
source decoder is forward error correction (FEC). For example, Annex H of
the H.263 standard provides an optional FEC mode. In this mode, 18 parity
bits are used to provide error detection and correction for each 493 video bits.
A commonly used method at the source decoder is the detection of syntax
and semantic violations. Examples of such violations are:
• An illegal codeword is detected.

• An invalid number of units is decoded. For example, the number of de-
coded DCT coe$cients within a block is invalid, the number of decoded
blocks within a MB is invalid, the number of decoded MBs within a
GOB is invalid, or the number of decoded GOBs within a frame is
invalid.

• A decoded motion vector points outside the permissible range.
• A decoded quantization parameter is out of range.
Another method that can be used at the source decoder is the detection of
violations to the general characteristics of natural video signals, for example,
the detection of strong discontinuities at the borders of blocks, blocks with
highly saturated colours (e.g., pink and green), or blocks where most pels
need clipping.
None of these methods guarantee nding all errors within a video bitstream.
In fact, the last method may sometimes detect an error-free block as an erro-
neous one. In practical systems, di erent combinations of these methods are
employed.
Having detected the occurrence of errors and identi ed their locations,
a number of methods can be used to combat the e ects of errors on the
video bitstream. The following three sections describe three categories of
error-resilience techniques: forward techniques, postprocessing techniques, and
interactive techniques. The three sections follow closely the classi cation used
in the comprehensive reviews by Wang et al. [179, 180].

9.6 Forward Techniques

In forward techniques, the encoder plays the primary role. Such techniques
work by adding a controlled amount of redundancy to the video bitstream.
This means that they sacri ce some coding e$ciency to gain in terms of error