摘  要
随着数字集成电路和计算机的飞速发展，以数字形式处理图象的需求越来越多。但是存储数字图象往往需要大量的比特位。这给传输、存储和处理图象带来了很大的困难。因此，如何对图象进行数据压缩是图象处理的关键问题。超大规模集成电路技术的发展，为各种复杂图象编码算法所需的巨大计算能力与低实现成本提供了可能，但有效地VLSI结构仍是提高图象编码系统性能的必需条件。根据图象编码算法及其VLSI结构的特点和发展情况，采用特殊设计的专用处理模块构成各种完整的编码系统，可获得较好的处理性能和灵活性。
矢量量化是一种有效的有损压缩技术，其突出优点是压缩比大且解码算法简单，因此已成为图象压缩编码的重要技术之一。矢量量化的关键技术是码书设计、码字搜索和码字索引分配。本文在静态图象压缩编码的应用背景下，重点研究矢量量化快速码字搜索算法以及它的硬件结构。一方面，分析现有的基于不等式判据和基于变换域的各种算法的优缺点，提出了一种新的基于排序哈德玛变换的快速码字搜索算法。另一方面，对已有的这些快速码字搜索算法进行仿真，分析它们的编码时间和计算复杂度，研究其硬件实现的可能性。此外，本文还提出了基于哈德玛变换快速码字搜索算法的硬件结构。
本文的主要目的是通过对矢量量化快速码字搜索算法的研究，改进已有的快速算法，进一步降低编码时间和计算复杂度，以获得更好的编码性能。通过对矢量量化快速码字搜索算法的VLSI结构的研究，为进一步设计更为详细和复杂的图象编码系统提供基础。
关键词  矢量量化；码字搜索；VLSI；图象编码

Abstract
With development of digital integrate circuits and computers, the requirement of image processing in digital form is increasing. But usually it needs a lot of bits to store up digital images. This makes it very difficult to transmit, store up and process images. Therefore, how to compress the data of images is the key to image processing. Advances in VLSI technology make it possible to implement the sophisticated image coding schemes with high processing power at a low cost. The high demands of image coding require special adapted VLSI architectures to achieve better performance. The approach of implementing hybrid image coding schemes base on basic function modules is attractive due to the best tradeoff between flexibility and performance.
Vector Quantization (VQ) is an efficient lossy compression technique, whose prominent virtues are high compression ratio and simple decoding process, so it has become one of important compression techniques for image coding. The three key techniques of basic VQ are codebook design, codeword search and codeword index assignment. In the application background of still image coding, this dissertation investigates mostly fast codeword search algorithms based on VQ and the VLSI architectures. On one hand, we analyse the virtues  and disadvantages of existing algorithms based on inequality criterion and transform domain, then a fast codeword search algorithm based on ordered Hadamard transform is presented. On the other hand, theses fast algorithms are simulated. We analysed their encoding time and calculating complexity, based on which we research the possibility of their architectures. In addition, the hardware architecture of a fast codeword search algorithm based on Hadamard transform has also been proposed.
The main purpose of this dissertation is to research the fast codewords search algorithms based on VQ, improve the existing fast algorithms and to reduce  encoding time and calculating complexity. Thereby we can get better encoding performance. Through the research of VLSI architecture of fast codewords search algorithms, we can provide a foundation for the implementation of detailed and sophisticated image coding schemes in the future.
Key words  Vector Quantization; Codeword Search; VLSI; Image Coding

目  录
摘要 I
Abstract II

第1章 绪论 1
1.1 课题的目的和意义 1
1.2 矢量量化原理介绍 2
1.2.1 矢量量化理论基础 3
1.2.2 矢量量化定义 4
1.2.3 矢量量化相关概念 5
1.3 矢量量化的关键技术 7
1.4 矢量量化码字搜索算法研究现状 8
1.5 矢量量化硬件结构研究现状 9
1.6 本文主要研究内容 10
第2章 基于不等式判据的快速码字搜索算法 12
2.1 引言 12
2.2 穷尽搜索算法 13
2.3 部分失真码字搜索算法 14
2.4 等均值最近邻码字搜索算法 15
2.4.1 均值不等式删除准则 15
2.4.2 等均值最近邻搜索算法 16
2.5 等均值等方差最近邻搜索算法 18
2.6 仿真实验 20
2.7 本章小结 21
第3章 基于变换域的快速码字搜索算法 22
3.1 引言 22
3.2 正交变换理论 22
3.3 基于小波变换的快速码字搜索算法 23
3.4 基于哈德玛变换的快速码字搜索算法 25
3.5 基于排序哈德玛变换的快速码字搜索算法 26
3.5.1 排序的哈德玛变换 26
3.5.2 码字删除不等式 27
3.5.3 算法步骤 28
3. 6 仿真实验 29
3.7 本章小结 30
第4章 快速码字搜索算法的硬件结构 31
4. 1 引言 31
4.2 穷尽搜索算法编码器的总体结构 31
4.3 基于和值不等式的快速算法的VLSI结构 33
4.3.1 基于和值不等式的快速算法 33
4.3.2 编码器的VLSI结构 34
4.4 基于哈德玛变换快速算法的硬件结构 37
4.5 各种快速算法计算复杂度的比较 40
4.6 本章小结 41
结论 42
参考文献 43
攻读硕士学位期间发表的论文 48
致谢 49

[1] [2]  下一页