基于MATLAB的图像融合算法

论文编号:TX244  论文字数:19816,页数:34

摘要
 
 图像融合能够将不同类型传感器获取的同一对象的图像数据进行空间配准。并且采用一定的算法将各图像数据所含的信息优势或互补性有机的结合起来产生新的图像数据。这种新数据具有描述所研究对象的较优化的信息表征，同单一信息源相比，能减少或抑制对被感知对象或环境解释中可能存在的多义性、不完全性、不确定性和误差，最大限度的利用各种信息源提供的信息。
    图像融合分为像素级、特征级、决策级三个层次，其中像素级图像融合能够提供其它层次上的融合处理所不具有的更丰富、更精确、更可靠的细节信息，有利于图像的进一步分析、处理和理解，它在整个图像融合技术中是最为复杂、实施难度最大的融合处理技术。本文的研究工作是围绕像素级图像融合展开的，针对像素级图像融合技术中需要解决的关键问题，研究了多种像素级图像融合方法。
 本论文的主要的研究内容有：
 首先介绍了图像信息融合的概念、优势、发展历史和应用领域，并介绍了图像融合的三个层次及常用的空域图像融合方法,空域融合方法有像素平均法、像素最大最小法、像素加权平均法，频域融合方法包括图像的多尺度分解、图像的小波变换、基于小波变换的图像融合方法。图像的预处理有滤波（邻域平均滤波法、中值滤波法）和图像配准。最后，对于图像融合系统来说，融合图像质量的评价显得特别重要，本文探讨了图像融合质量的评价问题，总结了融合效果的主、客观评价标准，作为本课题性能分析的判断标准。
 

关键词：图像配准；图像融合；空域融合法；小波变换；评价标准
 
 

MATLAB-based image fusion algorithm

Abstract
 
 The same object gotten from different sensors can be registered spatially by mage fusion. The information advantages or the complements of all the image data can be combined to produce new image data using some fusion algorithms. The new data can describe the optimized information of the studied object. Compared with single information source, the new data can reduce or restrain the ambiguity, the incompleteness, the uncertainty and the error, which may appears in the explanation of the studied object or the environment, and make full use of the information provided by all kinds of resources.
 Image fusion consists of such three levels as the Pixel level，the feature level and the decision level，among which the Pixel level image fusion can Provide more
abundant, accurate and reliable detailed information that doesn’t exist on the other levels and It is the most complicated in the whole image fusion techniques and also is the most difficult to implement in the fusion Processing techniques. this dissertation Progresses mainly around the Pixel level image fusion and proposes a variety of Pixel level image fusion techniques according to the key Problems in the Pixel level image fusion techniques.
 The major research and findings are as follows:
 First we introduce the concepts， advantages，developments and applications. Three levels of image fusion and image fusion techniques in common use are also reviewed. Airspace Image Fusion such as simple fusion method (pixel average, maximal or minimal pixel selection), Frequency-domain image fusion methods include the multiresolution image fusion techniques based on multi-scale pyramid decomposition, and the image fusion method based on wavelet transform Image Pre-processing like Filter processing (neighborhood average filter, median filtering method) and Image Registration. in the end, evaluation for fusion image is vital to fusion system. This dissertation probes into the image fusion quality assessment and deduces a set of indexes as the criteria to analyze the performances of this discussion.

Keywords: Image Registration；Image Fusion；Airspace integration method；Wavelet Transform；Evaluation criteria

目录
第一章 绪论 6
1.1 图像融合的概念 6
1.2图像融合的主要研究内容 7
1.2.1 图像融合的层次 7
1.2.2 图像融合算法的发展 9
1.2.3图像融合的步骤 9
1.3 图像融合技术的发展现状 10
1.4 本文的研究工作 10
第二章 图像预处理 11
2.1 图像的校正 11
2.2 图像滤波技术 11
2.2.1 邻域平均法 12
2.2.2 中值滤波 12
2.3 图像配准 13
2.3.1 图像配准概述 13
2.3.2 手动图像配准 14
2.3.3 基于图像特征的匹配算法 15
第三章 图像融合 16
3.1 加权平均融合法 16
3.2 像素灰度值选大/小融合方法 16
3.3 主分量融合法 17
3.4 IHS变换法 19
3.5 小波变换融合法 20
3.5.1 小波的定义及特点 20
3.5.2 基于小波变换的图像融合方法原理 25
3.5.3 图像融合规则及融合因子 26
第四章 图像融合效果评价 27
4.1 主观评价 27
4.2 客观评价 27
4.2.1 基于光谱特征的评价 27
4.2.2 基于信息量的评价 28
4.2.3 基于统计特性的评价 29
4.2.4 基于信噪比的评价 30
总结与展望 31
谢辞 32
参考文献 32