基于小波变换的图像融合算法研究

论文总字数：33871字

摘 要

图像融合是从信息融合发展而来的，利用原图像的信息，不仅能够保留大量的信息，而且也符合人们的视觉效果，合成的图像比任意一幅源图像所拥有的信息更丰富，更能够描述客观的现实。

小波分析被提出后，人们发现小波分析属于时频域分析，在时频两域都可以表示图像的局部特性。小波变换的图像融合方法可以聚集到图像的任意细节，所以常被称为数学上的显微镜。小波变换也被称为时间--尺度分析法，是一种多分辨率分析的方法。小波变换可以把图像分解为高频信息和低频信息，高频信息表示图像细节，低频信息表示图像轮廓。

本文研究的是基于小波变换的图像融合算法研究，提出了一种改进的融合算法，对传统的融合算法从三个方面进行了改进，包括融合规则，小波基和分解层数。本文对高低频信息的融合规则都做出了改进，低频信息采用加权平均的方法，通过区域能量和匹配度来确定低频信息所取的权。高频信息利用Cathy算子提取边缘，在边缘部分和其它部分采取不同的融合规则来达到凸显边缘的目的。实验中，熵、平均梯度、空间频率三个客观指标和主观效果都说明本文的融合方法更优。

关键字：图像融合、小波变换、区域能量

Research on image fusion algorithm based on Wavelet Transform

Abstract

Image fusion Developed based from information fusion ,it can make better use of information of original image, it can not only keep a lot of information, but also meet people's visual effect, the synthetic image have a richer information than any source images , more able to describe the objective reality.

After the wavelet analysis is proposed, it was found that belong to the time and frequency domain analysis, wavelet analysis in both time and frequency domain can represent the local features of the image.Wavelet transform image fusion method can gathered on any details of the image, so often referred to mathematical microscope.Wavelet transform is a kind of signal time - scale analysis method, has the characteristics of multi-resolution analysis.Wavelet transform can put the image decomposed into high frequency and low frequency information, high frequency information represents image detail, low frequency information represents the image contour.

This article studies the image fusion algorithms based on wavelet transform, presents an improved fusion algorithm, the traditional fusion algorithm is improved in three aspects, including fusion rule, wavelet and decomposition layers.In this article,the high frequency and low-frequency information fusion rules are improved, low-frequency information using the weighted average method, the weight was identified through the regional energy and matching rate.High frequency use Cathy operator to extract the edge information, at the edges and other parts adopt different fusion rules to achieve the purpose of highlight edge.In experiment, three objective indicators include entropy, average gradient, spatial frequency and subjective effect suggests that the method of this article has a better result.

Key words: image fusion, wavelet transform and regional energy

目 录

摘要.........................................................................Ⅰ

Abstract.....................................................................Ⅱ

第一章 绪论...................................................................1

1.1本课题研究的背景和意义................................................1

1.1.1背景............................................................1

1.1.2意义............................................................1

1.2 图像融合概述..........................................................1

1.3 图像融合应用..........................................................2

1.4 国内外现状............................................................3

1.5 本文的主要内容........................................................3

第二章 小波理论...............................................................5

2.1 小波变换..............................................................5

2.1.1介绍.............................................................5

2.1.2 连续小波变换.....................................................6

2.1.3 离散小波变换.....................................................6

2.2 多分辨率分析与小波变换快速算法........................................7

2.2.1 多分辨率分析....................................................7

2.2.2尺度函数和尺度空间 .............................................7

2.2.3 小波变换快速算法................................................8

2.3 几种常用的小波基函数..................................................9

2.4 Mallat的快速算法.....................................................11

第三章 图像融合方法..........................................................12

3.1 传统的融合方法........................................................12

3.1.1 绝对值最大......................................................12

3.1.2 区域最大........................................................12

3.1.3 区域能量最大....................................................12

3.1.4 邻域和最大......................................................13

3.2 改进的区域能量与低频融合规则...........................................13

3.3 边缘算子与高频融合规则.................................................15

3.3.1 边缘算子........................................................15

3.3.2 高频系数融合规则................................................15

3.4 本文提出的融合方法.....................................................16

3.5 图像质量评价...........................................................16

3.5.1 图像质量的主观评价...............................................16

3.5.2 图像质量的客观评价...............................................17

第四章 实验研究过程..........................................................19

4.1 多传感器图像融合.......................................................19

4.2 可见光与红外光图像融合.................................................28

4.3 医学图像融合...........................................................32

第五章 总结和展望............................................................36

致谢.........................................................................37

参考文献（References）.......................................................38

附录.........................................................................39

剩余内容已隐藏，请支付后下载全文，论文总字数：33871字