In this work, our focus is on video streaming applications with relatively strict delay constraints; for such applications, forward error correction (FEC) is the preferred channel coding technique to recover from packet losses. Moreover, since video packets are usually of different importance, optimal bit allocation across video packets results in different packets receiving unequal error protection (UEP). For Internet video, cross-packet FEC in addition to source packets can be performed by generating parity packets. Alternatively, a source packet can be partitioned into different transport packets to achieve prioritized protection, which has been studied for scalable video in [2]. In this paper, we compare the performance of these two packetization schemes applied to single-layer video. The comparison is carried out in a joint source-channel coding (JSCC) framework, where error resilient source coding and FEC are jointly designed. We show that either packetization scheme may be optimal depending on the packet loss rate.
展开▼