多视点视频编码中的资源分配与目标决策相关理论研究

An increasing demand is currently witnessed for Multiview Video System since it provides the novel 3D visual perception and interactivity enjoyment, which are useful for many applications. Due to the large volume of multiview video data, the high compression performance and low complexity requirements must be considered during coding stage. Traditional Multiview video coding (MVC) method try to reduce the temporal, spatial, inter-view, frequency and visual redundancy based on information theory or psychophysics. However, due to large data volume, higher compression efficiency and lower complexity of MVC algorithm are still desired. For higher MVC compression performance, the key optimization problems, mainly resource allocation and decision making problem, shall be solved. This project will firstly analyze and model the basic optimization problems in MVC, and then investigate the resource allocation and decision making optimization theory, respectively. Based on these theories and basic solutions, the optimization theory based MVC method will be investigated by joint optimization with the predictive coding theory and perceptual coding, including bit allocation optimization in MVC, multi-level and multi-object decision making based MVC method and multiview video plus depth map coding. This project will establish a framework of optimization theory for multiview video and achieve the theoretical innovation and technological breakthroughs for MVC, which will promote MVC performance and the development of the 3D video signal processing related research fields.

多视点视频系统因为能提供立体感和交互性的全新视觉体验，具有广泛的应用前景，然而海量多视点视频的高效压缩和有效计算问题是其推广应用急需解决的关键性难点问题。传统多视点视频编码方法基于信息理论和视觉感知，消除时域、空域、视点间、频率和视觉冗余。为在此基础上进一步提高视频压缩效率和效果，本项目将首先模型化分析多视点编码中的两个重要优化问题，即有限资源最佳分配和目标决策问题，并基于博弈论与最优停止理论研究这两个问题的一般性最优求解与相关优化理论。然后，以此优化理论为基础，结合编码信息论中的预测编码与感知压缩特性，研究基于优化理论的高效多视点视频编码理论与方法，具体提出基于优化理论的多视点码率分配和方法，基于多层次目标决策与优化的多视点编码方法和彩色与深度联合编码等方法。本项目探索具有普遍性的多视点视频优化理论，实现编码优化的理论创新与技术突破，促进三维视频信号处理领域的研究与发展。

本项目的研究目标为围绕多视点视频编码中存在的资源分配和目标决策两类最优化问题，研究多视点视频编码中的优化问题表示以及有限资源条件下的资源分配和目标决策相关理论。并在此基础上，研究基于优化理论的多视点视频编码优化方法。在多视点视频编码优化方面实现了核心理论与关键方法上一定的突破，具有较高的原创性。本项目执行期间的成果包括学术成果、人才培养、国际影响等几个方面。其中的学术成果包括（1）学术论文36篇，其中国际SCI期刊32篇（国际权威期刊IEEE Transactions系列刊物23篇），国际EI会议4篇（其中CCF推荐的B类会议1篇，C类会议1篇）；（2）国际学术会议4次；（3）申请美国发明专利2项（授权1项）。在人才培养方面，项目组负责人于2014年当选IEEE Fellow，项目组成员赵铁松博士，2016入选中组部“青年千人”计划，在福州大学晋升教授；项目组成员王旭、曹菁菁、王冉、李珂和潘兆庆获得香港城市大学博士学位，并获得国内外知名大学教职。 在国际影响方面，项目负责人邝得互目前担任IEEE Transactions on Industrial Electronics (2009-)、IEEE Transactions on Industrial Informatics (2007-)和Information Science (2011-)等国际权威期刊副编辑，担任IEEE SMC协会会务分管副主席等职务。