The postprocessing error concealment techniques presented in the previous section are adequate for providing an acceptable video quality at the decoder only when the packet loss rate is very low. These techniques perform much better when they are used in conjunction with forward error control techniques. Obviously, the interaction between the encoder and the decoder should take place over a two-way communications channel. This prerequisite cannot be satisfied for real-time video applications in specific network architectures, such as broadcast or multicast channels, as the backward requests from the decoders would place a great burden on the network, further increasing packet loss.4.3
For real-time video communications, plain buffering and retransmission are not applicable due to the latencies incurred. One of the simplest approaches is that the decoder acknowledges the loss of specific packets to the encoder and requests the transmission of an intra-coded frame to stop error propagation. This corrects the error for the future frames, after a time period which can be considerable if the round-trip delay is large. However, it is not economical, as the low-compression-efficiency intra-coding might be invoked for every -- small or large -- quality degradation due to packet losses. Furthermore, low-bandwidth video coding standards such as H.263 are sending whole intra-coded frames on very rare occasions during the video sequence.
More sophisticated schemes are sending back information on the specific region of the picture that has not been correctly received, and the encoder intra-codes and sends only the corresponding macroblocks. These selective encoding schemes have been further elaborated by using techniques such as error tracking. The idea behind error tracking is to simulate at the encoder the loss that the decoder experiences, according to the feedback information it receives. Then, motion estimation is applied to the locally reconstructed error-struck images, and the updated motion vectors and prediction errors are sent as in the normal, compression efficient, interframe mode. On their reception, the receiver can effectively reconstruct the new correct video frames.
Other solutions include H.263's reference picture selection mode, in which both the encoder and the decoder have multiple frame buffers, which are selectively used depending on the nature of the messages exchanged between them. It comprises sets of coding parameters which trade compression efficiency for error resilience when the loss rate is high, and vice-versa.
Modern Internet video streaming applications use techniques such as prioritised multicopy retransmission [WZ98]. As their name suggests, these schemes combine prioritised layered coding, with multiple channel retransmission for the base layer. Since the transit delay is not crucial for one-way video communications, these techniques provide flexible tradeoff between delay and reconstruction quality.