Multimedia coding and communication for network applications

多媒体流跨越了广泛的网络应用，从视频会议和IP语音（VoIP）到互联网电视广播（IP-TV）。据目前的估计，流媒体占整个互联网流量的四分之一以上。因此，媒体流问题的有效解决方案有可能直接影响互联网的带宽经济。当前的流媒体解决方案基本上将网络媒体通信分为两层：媒体压缩层和网络通信层。相比之下，拟议的研究旨在开发超越当前分层方法的新解决方案。这些新解决方案的开发是基于这样一种认识，即当媒体处理和网络交付方法分开设计时，存在固有的性能损失。通过联合考虑源压缩和网络通信，我们开发的解决方案将能够更有效地利用可用的网络基础设施和带宽资源。为了实现这一潜力，将开发一个优化框架，以捕获源编码和网络通信的复杂纠缠。这种优化的目标将是在给定底层网络的带宽限制的情况下，最大化对所有网络客户机的服务质量的集体度量。在此框架内，将研究良好的网络通信策略（如路由或网络编码）和源编码元素（如渐进式编码或更通用的多重描述编码）进行联合优化。简单的例子（如提案中报告的例子）表明，使用这些新方法进行数字多媒体分发的影响可能是显著的。

Multimedia streaming spans a wide range of network applications, from video conferencing and voice over IP(VoIP) to TV broadcasting over the Internet (IP-TV). Streaming, as current estimates suggest, accounts for more than a quarter of the overall Internet traffic. As such, efficient solutions to media streaming problem have the potential to directly impact the Internet's bandwidth economy. Current streaming solutions essentially separate network media communication into two layers: media compression layer and network communication layer. In contrast, the proposed research aims at developing new solutions that go beyond the current, layered, approaches. The development of these new solutions is based on the realization that there is an inherent loss of performance when media processing and network delivery methods are designed separately. By a joint consideration of source compression and network communication, our developed solutions will be able to utilize available network infrastructures and bandwidth resources more efficiently. To realize this potential, an optimization framework will be developed that captures the complex entanglement of source coding and network communication. The objective of such optimization will be to maximize a collective measure of quality of service to all network clients, given the bandwidth constraints of the underlying network. Well studied network communication strategies (e.g., routing or network coding) and source coding elements (e.g., progressive coding or more general multiple description coding) will be jointly optimized within this framework. Simple examples (such as the one reported in te proposal) suggest that the impact of using these new approaches to digital multimedia distribution can be significant.