基于矢量量化的零树小波图像压缩方法-应用数学专业论文.docxVIP

摘 要 针对矢量量化方法往往存在的分块效应和 EZW 算法在高压缩比情况恢复的图像质量 较差的问题，是否能加入某种模块而将矢量量化、EZW 联系起来，寻找出一种低比特率高 信噪比的压缩算法值得研究。 本文首先基于 SOFM-C 算法引入辅助神经元(Auxiliary neurons)的概念提出了带有良心 的辅助神经元自组织映射(ASOFM-C)算法，不仅拥有在码字利用更为均衡的优点，同时保 持了该神经元的特性。然后针对 ASOFM-C 与 SOFM 神经网络失真测度定义的不同，提出 适用于 ASOFM-C 的快速搜索获胜神经元的加速定理。最后本文提出一种基于矢量量化的 嵌入式零树小波方法，基本原理是引入差值图像思想，差值图像经过小波分解以后会存在 大量为 0 的小波系数，再利用 EZW 编码帮助提高重建图像质量。实验表明，相对于 EZW 和 JPEG2000 算法，本文算法的压缩比和编码质量均有显著提高。 关键词：图像压缩；矢量量化；Kohonen 神经网络 ；EZW - I - Abstract Vector quantization method for block effects are often present in the EZW algorithm and high compression ratio of the image restoration problem of poor quality, it will be adding a module and vector quantization. EZW linking to find out a low bit rate signal to noise ratio of the compression algorithm is worth studying. Based on SOFM-C algorithm ,we introduced the concept of auxiliary neurons (Auxiliary neurons) to a new self-organizing map (ASOFM-C) algorithm.It is not only a code word in a more balanced use of the advantages,and maintaining the characteristics of neurons. Based on the SOFM-C and SOFM neural network have a different distortion measure.A fast search neural acceleration theorem is proposed for the SOFM-C .This paper presents a vector quantization based on the embedded zerotree wavelet method, and its basic principle is the introduction of the difference image of thought.At first the original image vector quantization, then the original image and restored image vector quantization of the difference image requirements. Poor binary image after wavelet decomposition later .There are a lot of wavelet zero coefficients to help improve reconstructed image quality. Experiments show that, compared to EZW and JPEG2000 algorithm, the algorithm of compression ratio and coding quality have improved significantly Key Words：Image impression；Vectorquantization；Kohonen neural network；EZW 目 录 摘 要 I Abstract II 1 绪论 1 1.1 本文选题背景及主要内容2 1.1.1 本文选题背景 2 1.1.2 主要内容 2 1.2 图像压缩背景知识 3 1.2.1 图像压缩分类 3 1.2.2 图像编解码模型 4 1.2.