面向WMSN的分布式多视点视频编码及鲁棒传输关键技术研究

With the availability of low-cost hardware, Wireless Multimedia Sensor Networks(WMSN) have been potentially applicable to several emerging applications. The project will be mainly focused on the problems of video coding and transmission over error prone channels. The main objective of this project is to develop new frameworks, algorithms and paradigms for low power consumption, low delay, high coding efficiency and error resilient video communication systems, specifically for practical WMSN. .The main tasks of our project are as follows:..Task 1.Symmetric Distributed Multi-Source Coding (SDMSC) based on Slepian-Wolf-Cover network theorem. Good DSC system need to be designed to achieve the entire Slepian-Wolf-Cover region, which is required for the following Distributed Video Coding (DVC) scenarios : (1) Rate balancing among different views and different time instants of each view.(2) Flexible rate allocation of each view. In this task we will mainly concentrate on three-sources' symmetric DMSC based on source splitting technology and rateless channel coding...Task 2. Symmetric Distributed Multiview Video Coding (SDMVC) based on symmetric distributed source coding. In this task we will take three temporal and spatial continuous video frames as basic elements and develop a new symmetric DMVC system to achieve flexible rate allocation among views. Iterative hybrid framework will be developed to exploit both temporal and spatial correlation of multi-view video sources, which will improve coding efficiency of the whole system. ..Task 3.Distributed Joint Source-Channel Coding (DJSCC) based on rate-adaptive channel coding and its application in distributed video coding. In order to simultaneously achieve high coding efficiency and error resilience under the memory restriction of WMSN, DJSCC based on single channel code must be developed to correct errors in both the virtual channel and the real transmission channel. At the same time, 'Ratelessness' is also required to cope with time varied virtual channels and real transmission channels. In this task we will develop DJSCC based on rate adaptive channel coding, which will consider both virtual channel and real transmission channel's parameters...Task 4.Encoder-rate-allocation based on joint source-channel parameter estimation and data packetization scheme based on several strategies. Rate allocation before transmission is vital for low delay DVC, which will directly allocate enough parity data for transmission without feedback channel. On the other hand, how to packetize the allocated data can also impact system's error resilience. So in this task we will firstly develop efficient encoder-rate-allocation algorithms for low delay DVC, upon which new packetization paradigms will be developed for better error resilience. Special attention will be paid to reconstruction functions of DVC for different packetization paradigms.

随着无线多媒体传感器网络(WMSN)在各种视频领域的广泛运用，满足摄像头低计算复杂度、低存储能力等前提下对数据进行高效压缩和高鲁棒性传输成为迫切需要解决的问题。分布式多视点视频编码(DMVC)技术具有解决上述问题的巨大潜力。当前国内外对DMVC的关键技术研究还处于初级阶段，研究成果大都以各种理想条件为前提，并很少考虑在无线网络中的抗误码传输问题。本项目拟结合WMSN特点，研究DMVC及鲁棒传输关键技术。主要创新点：1.提出基于Slepian-Wolf-Cover网络理论的对称多信源编码模型以实现分布式信源编码中各信源传输码率均衡和同步解码。2.提出基于对称多信源编码模型和分层迭代模型的时空域码率均衡DMVC系统。3.提出基于码率自适应信道码的分布式联合信源信道编码系统，实现对视频的分布式编码及鲁棒性传输。4.提出基于信源信道联合参数的码率估计算法，并对数据最优化打包进一步提高传输鲁棒性。

针对无线多媒体传感器网络(WMSN)中的视频数据压缩及传输问题，研究低编码复杂度、高压缩率、高误码鲁棒性的视频通信技术具有非常重要的基础研究意义及实际应用价值。分布式多视点视频编码(DMVC)只需在解码端完成视点内、视点间冗余信息的挖掘，就能在整体上实现视频数据的高效编码，这为WMSN提供了理论解决方案。然而DMVC技术要从真正意义上应用于WMSN需要解决多个理论问题，包括压缩效率、视点间及视点内码率均衡、误码鲁棒性、系统编解码实时性等。本项目主要针对这些问题开展了深入研究。.传统基于不对称结构的DMVC技术存在的视点间及视点内码率不均衡问题，对无线信道传输的稳定性带来了挑战。针对这个问题本项目深入研究了分布式对称多信源编码理论，实现了码率对称的多信源分布式编码。在此基础上提出了基于对称结构的DMVC结构框架，实现了DMVC视点间及视点内的码率均衡。.针对DMVC系统的压缩效率，本项目着重从边信息生成、量化模型、虚拟信道模型和重建函数等方面进行了深入研究，有效地提高了系统的压缩效率。.针对DMVC系统抗误码传输，项目首先从信道码入手，研究了基于码率自适应信道码的分布式信源信道联合编码（JSCC）系统。然后通过提出多描述DVC系统、提出新的码率控制及打包机制、提出针对误码环境下Wyner-Ziv(WZ)帧的优化重建算法、提出基于WZ码流保护的关键帧抗误码方案等来提高系统整体的传输性能。.针对WMSN环境下视频传输的实时性问题，本项目从理论算法及应用设计两方面开展了深入研究工作。项目首先针对编码端关键帧编码问题，对帧内编码的实时性提升技术开展了研究。然后针对解码端实时性问题，从DMVC框架下边信息生成快速算法、HEVC框架下快速运动估计/补偿算法、基于FPGA/DSP硬件的信道码及视频处理加速等方面开展了深入研究。最终为WMSN环境下的DMVC系统实时性的提高提供了有效的解决方案。.总的来说，本项目面向WMSN系统对DMVC的压缩及鲁棒性传输关键问题进行了全面深入的研究，已发表和录用论文28篇，其中SCI论文10篇，EI论文4篇，研究内容无调整和变动，达到了预期目标，从理论及应用技术方面为WMSN中的视频压缩及传输问题提供了有效的解决方案，既有重要的基础研究意义，又有潜在的实际应用价值。