As seen in previous discussion on the pros and cons of slice-coding, a clear optimal number of slices per video frame for NAL packetization cannot be determined. [4] and [7] claim that reasonable number of packets/slices per video frame is around 10, and the resulting packet size in this case is in the range of 100 bytes. In packet-lossy transmission over wired networks, it is all right to employ such fixed 10-slice NAL
Chapter 4 Adaptive H.264/AVC Network Abstraction Layer Packetization packetization to have constant packet payload size because the wired channel is not time-varying and the error rate is sufficient low so that reasonable network throughput and system efficiency can be achieved. However, in wireless networks, as the channel changes with time, it may be hostile at the moment, but it may also be improved in the following period. Figure 4.7 shows the channel status in terms of block error rate (BLER) and BER for JVT test conditions error pattern 1, which characterized with mobility3km/h and average BER9.3×10−3. The BLER is defined such that as long as there is at least one bit in the block is erroneous, the whole block is declared as corrupt.
0.00001 0.0001 0.001 0.01 0.1 1
0 50 100 150 200
Frame number
Probability of error
BLER BER
Figure 4.7: Time-varying channel status
It can be seen that the channel is continuously varying with time in the long run. In the short run, it varies slowly. And BLER generally follows the BER. The discontinuity in the figure shows that the wireless channel is error free at that particular moment, which cannot be plotted on logarithmic scale. Such time-varying nature of wireless channel means that during the error free period, more source data can be sent and therefore, scarce wireless bandwidth can be fully utilized. However, with such fixed payload size for video data regardless of channel condition, wireless channel cannot be fully utilized. In addition, this fixed NAL packetization approach without
Chapter 4 Adaptive H.264/AVC Network Abstraction Layer Packetization
52 considering channel conditions will also affect the redundancies added by lower layer FEC and ARQ if error control techniques are adopted. In other words, it is possible that the system runs into the situation that there is less channel protection for source video data when the channel is hostile, while the source video data is overly protected when channel is amiable. Such inflexibility will lead to low throughput, high transmitting power, and less efficiency of video transmission system.
Unfortunately, recent research on H.264/AVC [25-27] does not address the issues on the NAL packetization to enhance error resilience and system efficiency, fixed NAL packetizaiton scheme, more precisely, NAL packetization with fixed slice partition is adopted without considering wireless channel conditions [66]. Nevertheless, since it is possible to estimate the channel behavior for next short period based on current channel status, above observations motivate the possibility of H.264/AVC NAL packetization with slice partition adaptively to wireless channel conditions. In addition, Stockhammer et al. point out that it is worth noting that new directions in the design of wireless systems do not necessarily attempt to minimize the error rates in the system but to maximize the throughput [7]. This is especially appealing for services with relaxed delay constraints and certain error tolerance in the end-user quality.
Therefore, the motivations of proposing a novel adaptive H.264/AVC NAL packetization scheme can be summarized as follows:
i) To take advantage of slice-coding in assisting error control techniques by localizing the burst errors occurred in wireless environment so that the end-user quality can be improved with the assistance of error-concealment techniques;
ii) To facilitate throughput adaptation in time-varying wireless environment so that the network or system efficiency can be improved in conjunction with lower layer error control mechanisms under cross layer optimization.
Chapter 4 Adaptive H.264/AVC Network Abstraction Layer Packetization More specifically, the novel adaptive H.264/AVC NAL packetization scheme consists of adaptive slice partition and “simple packetization” for the partitioned slices. When wireless channel is hostile, large number of slices per video frame is preferred to assist error control techniques such that heavy channel protection may not be necessary.
When the channel is amiable, smaller number of slices per video frame is preferred to avoid unnecessary overheads from network protocol headers. By doing NAL packetization in such adaptive way, system efficiency can be improved.