Compressed video is extremely vulnerable to transmission errors, as discussed in previous chapters. First, in intra-coded pictures, a packet usually holds a good deal of information, spanning over a region of the picture. Loss of the packet causes the whole region to be unavailable to the decoder. The results aggravate if appropriate measures, for example resynchronisation headers and location identifiers, are not used to prevent loss of synchronisation at the decoder leading to spatial propagation of the error.
Second and last, interframe prediction is causing spatiotemporal propagation of
potential errors, as depicted in Figure 4.1 for the case where a
H.263 GOB is erroneously received. The error decays over time
due to the leakage in the interframe prediction loop, see also
Figure 2.6, page ![[*]](cross_ref_motif.gif)
. Intra-coded frames
terminate the effect of such errors, but, in practical low-bandwidth
transmission schemes such as H.263, they are inserted in the sequence only after
a large number of inter-coded frames. In such cases techniques as conditional
replenishment are used, as described in the following sections.
