Conclusions

The target in this work is a reliable digital video transmission scheme over IP-based networks. The latter, as non-reliable best-effort delivery systems, are prone to packet loss, as described in Chapter 3. Moreover, compressed data exhibit very high sensitivity to partial loss, due to the concentration of significant information to the resulting bitstream. This is clearly visible in video sequences, where, due to the predictive compression techniques used, errors are propagated in the spatial and temporal direction, leading to very low quality, nearly imperceptible, images. Therefore, the plain standardised compression algorithms used in H.261, H.263, and MPEG codecs, are shown not to perform well in such conditions. Some solutions sacrifice much of the compression efficiency of these schemes, by disabling interframe prediction, in order to provide a more robust, self-synchronising bitstream. This is arguably acceptable, as tradeoffs between compression efficiency and error-resilience are common practice. However, the price of sacrificing all interframe prediction is very high, as bandwidth requirements are significantly and wastefully increased. Recent advances in research on coding and transmission schemes, target on retaining the compression efficiency of interframe prediction and using judiciously intraframe synchronisation to increase robustness in packet loss conditions. The most sophisticated techniques involve cooperation of the sender and the receiver in order to recover from packet loss. Apart from enhancing the reliability and robustness of the transmitted information, it is desirable that quality degradation does not reach levels frustrating for the users, at any time. Therefore, besides the encoding, error concealment techniques are applied at the decoder, to increase the perceptibility of error-struck parts of the video sequence. It is well understood that these techniques are only enhancement measures, and do not remedy the information loss effects. As in most engineering design problems, there is not a panacea providing solutions for any application. Robust video transmission schemes are limited scope systems, optimised and orientated to particular application requirements. For this work, these requirements were set for personal communications applications over the current unreliable IP network service model. Taking into account all the above considerations, the resulting proposed solution was to insert a limited amount of redundancy in the waveform encoding stage, by slightly modifying the intra-inter coding decision rule to be more biased than usual towards intra-coding. This provides the desired robustness to the bitstream and a good reference to the decoder to fully recover from errors in only a few fractions of a second. The performance of the scheme is further improved if a simple temporary error repair techniques, such as temporal prediction is used, as shown by the experimental results obtained through simulation. The proposed scheme copes well with prolonged error bursts due to congestion, as intra-coding provides effective isolation of groups of frames, making their decoding independent of the correct reception of other groups. It does not require two-way communication between the sender and the receiver, which could make it a good candidate for multicasting. Its main drawback is the limited compression efficiency associated with intra-coding. In addition, it does not perform well under unfavourable conditions such as packet loss on contiguous intra-coded packets, which have a very low, but non-zero occurrence probability. Future research should focus on interactive error control and transmission, which undoubtedly provides the best solution for robustness. The very simplistic network model used, with random packet loss and no QoS guarantees, could also be replaced by a more near to real world networks simulation model. Furthermore, efficient ways of handling errors of a multicast group during a session, could be investigated as well.

Hosted by www.Geocities.ws