【作者】 纪中伟；

【导师】 朱维乐；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2002， 博士

【摘要】 本文自始至终贯穿着运动估计和运动补偿的思想，在此思想的指导下，通过对视频图像序列的时空相关性分析，对运动估计和运动补偿进行算法研究。并将本文提出的先进的运动估计和补偿算法应用到去隔行显示器增强算法和视频编码等数字视频处理领域中。第二章首先详细讨论了运动估计的基本理论，特别研究了基于块的运动估计方法，包括块匹配模型、块匹配原理、匹配准则和搜索过程算法等等。然后基于预测性运动矢量的概念，利用相邻块运动矢量的相关性和全局极值点概率分布特性，提出了预测性菱形搜索算法和它的改进算法：自适应预测性菱形搜索法，设计出具体算法，并给出了与传统快速块匹配法比较的计算机仿真结果。第三章首先详细阐述了运动补偿的基本理论和传统运动补偿的基本方法，然后讨论了多假设运动补偿原理与重叠块运动补偿原理。最后在此基础上根据自适应运动建模的思想，引入了本文所提出的自适应重叠块运动补偿方法。此外还讨论了与运动补偿密切相关的运动补偿后处理中常用的滤波原理和方法。前面两章只讨论了作者提出的运动估计和运动补偿方法的原理，从第四章开始我们将把它们运用到实际的数字视频处理应用中去。第四章讨论了运动估计和运动补偿在去隔行显示器增强算法中的应用，考虑了场间的运动信息，将前两章中介绍的预测性菱形搜索和多假设运动补偿原理运用到实际的去隔行算法中。创新性地提出一种快速去隔行技术，并给出了方案设计、软件框图和实际仿真结果。第五章讨论了本文运动估计和运动补偿方法在视频编码中的应用，首先阐述了视频编码理论和编码系统各个子模块的基本原理，然后在预测性菱形搜索和自适应重叠块运动补偿基础上运用变换编码加运动补偿的预测编码方法实现一个混合视频编码的方案。并给出了具体的算法实现和仿真结果。第六章讨论了本文运动估计和运动补偿方法在小波域视频编码上的应用，在阐述了小波基本理论和小波系数分类量化编码基本原理之后，将上一章运用的离散余弦变换用离散小波变换替代，然后运用小波变换编码和基于自适应重叠块运动补偿的预测编码方法实现一个混合视频编码的方案。并给出了具体的算法实现和仿真结果。更多还原

【Abstract】 This dissertation is guided by the idea of motion estimation and compensation. Based on this idea, the advanced algorithm about motion estimation and compensation are developed using the spatial and temporal correlation in video sequence. At the same time, the application of the discussed methods in de-interlacing and video coding is also given. Chapter 2 first reviews all kinds of basic theory concerned by motion estimation (ME), especially ME based on block matching, which includes block matching model, block matching theory, matching rule, searching processing and so on. Then based on the idea of predictive motion vector, using of spatial correlation of adjacent block and global minimum points probability distribution characteristic, predictive diamond searching (PDS) and its advanced mode: adaptive PDS (APDS) are introduced. Finally the algorithm of PDS and APDS and its simulation results comparing with conventional ME algorithm are given. Chapter 3 reviews all kinds of basic theory and conventional algorithm concerned by motion compensation (MC). Then the multihypothesis MC (MMC) and overlapped block MC (OBMC) are discussed. Finally, adaptive overlapped block MC (AOBMC) on the basis of adaptive motion modeling is introduced. Furthermore, the practical filtering theory and algorithm correlated with MC is reviewed too.The former two chapters only describe theoretic ME and MC algorithm. From chapter 4 on, the theories are applied to practical digital video processing. The application of proposed algorithm in de-interlacing is discussed in chapter 4. Considering the motion information and correlation of spatial motion vector field, theory about PDS and MMC is applied to de-interlacing algorithm. Then, together with implementing the algorithm, software structure and its simulation are provided. The application of proposed algorithm in video coding is discussed in chapter 5. By first reviewing the basic video coding theory and then analyzing kinds of sub-module in video coding system. Adaptive overlapped block motion compensation (AOBMC) is combined with PDS technology for temporal prediction. The system composed of hybrid transform coding and motion compensation prediction coding is constructed. Finally corresponding algorithm and simulation are provided.Chapter 6 introduces the application of proposed algorithm in wavelet video coding. After discussing basic wavelet theory and the technique of wavelet coefficients classification and quantization, we apply discrete wavelet transform instead of discrete cosine transform. The proposed<WP=7>method is the hybrid wavelet motion compensation video coding, which combines adaptive overlapped block motion compensation for temporal prediction with the zero-tree wavelet technology for spatial decorrelation. Corresponding algorithm and simulation are given at last.更多还原