基于BP神经网络图像压缩研究.pdf

摘要
基于神经网络的图像压缩技术在理论和技术上开辟了图像压缩的
新途径。本文在深入研究基于 BP(Back-Propagation)网络的图像压缩方
法之后,针对标准 BP(Back-Propagation)网络需要较长训练时间和完全
不能训练的缺点,提出了联合改进方法,并由此引出了神经网络图像
压缩系统的研究与探讨,使其技术本身更具应用前景。
归纳起来,本文主要围绕以下几个层次进行展开:
(1) 首先从探讨 BP 算法入手,剖析基于 BP 网络的图像压缩机理,
深入研究它在图像压缩中的应用及其关键技术,通过一系列实验,分
析并总结了压缩性能与各种网络参数之间的关系,这是本论文工作的
一个重要部分。
(2)在 BP 网络实现图像压缩的基础上,针对标准 BP
(Back-Propagation)网络需要较长的训练时间和完全不能训练的缺点,
提出了联合改进方法,即在转移函数中引入陡度因子和各层权值调整
变尺度的两个方法。这是新的尝试,也是本文的创新之一。
( 3 ) 研究小波神经网络的算法及其结构, 重点阐述了辅助式的结
合方式一小波神经网络,并尝试将小波和改进后的 BP 神经网络相结
合,实现图像压缩,最后分析了各种参数对重建图像性能的影响,这
是本论文工作的另一个创新之处。

关键词:图像压缩;BP 神经网络;小波变换;小波神经网络
Abstract
pression based on artificial work provides a novel way for
investigation of theory as well as technique in the field of this
paper,We discuss deeply pression based on back-propagation neural
network and anization feature ,we propose a modified
vector quantization method aiming at the deficiency of original SOFM VQ,which not
only demonstrates the potentials of the proposed methods,but also provides a
profound insight into the theory of pression based on artificial neural
network.
In sum,this paper is developed according to the following hierarchy and research
mechanism.
(1)BP work can provide the ability of pression we
firstly discuss BP algorithm and explode the mechanism of pression based
on BP we make a study of the key technology in the application
and several kinds of learning rules are used press the static a
series of experiments are executed,we analyze and summarize the relationship
between pression performance and the parameters of BP
is a main part of this paper.
(2)bination with the correlativity between image blocks,we propose the
idea of pression based on hierarchical BP further
analysis is conducted in work model and the nested training
pression and image reconstruction are plished
method