【作者】 赵天昀；

【导师】 李正文；

【作者基本信息】 成都理工大学 ， 信号与信息处理， 2004， 硕士

【摘要】 图像融合是一种信息综合处理技术，其主要目的是通过对多幅图像间冗余数据的处理提高图像的可靠性，通过对多幅图像间互补信息的处理提高图像的清晰度。近年来，图像融合已成为图像理解和计算机视觉领域中一项重要而有用的新技术。本文主要探讨像素级图像融合技术。 文中首先介绍了图像融合技术研究的背景，描述了图像融合的基本概念、层次、应用及国内外研究动态，阐述了像素级图像融合技术的基本原理和方法，并提出了一种对融合方法性能及融合图像效果进行主、客观评价相结合的评测方法，通过后文的大量实验证明该方法是合理有效的。 在基于空域的图像融合方法一章中，首先介绍了简单融合方法，在加权平均融合方法的基础上，提出利用PCA变换确定加权平均系数，并通过实验证明利用该方法可得到融合效果更佳的图像。随后还介绍了总体性能优于简单融合方法的基于区域特征的融合方法。 论文第四章重点介绍基于小波变换的图像融合方法。在介绍了小波变换的基本理论、多分辨率分析的思想、图像的小波变换及其Mallat快速算法之后，讨论了基于像素和基于区域的融合规则及实现方式，并提出新的融合规则。通过一系列图像融合实验，利用文中所提出的评测方法研究了不同小波基、不同小波分解层数、不同融合规则、不同特性区域大小、不同阈值对融合性能的影响，从而得到相关结论。 本文最后提出一种基于图像分割块的新融合方法。通过实验证明该方法能够准确保持有用信息，实现信息互补，具有计算较简单、处理时间较短，融合效果较好的优点。图像融合技术可应用于多个领域，本文还对它在医学领域的应用做了探讨，也进一步验证了文中的各融合方法。更多还原

【Abstract】 Image fusion is an integrated information processing technology, which mainly aims to improve the reliability of images by processing the redundant information and the definition of images by processing complementary information among several images. In recent years, image fusion has became into a new important and useful technology in the field of image analysis and computer vision. This paper is mainly to discuss pixel-level image fusion technology.After introducing the background of research, basic concepts, development, internal and external application and researches of image fusion technology, the paper elaborates the fundamental principles and methods employed in pixel-level image fusion technology and puts forth a combination of the subjective and objective evaluation criteria for image fusion performance and the effect of the fused images, which is proved to be reasonable and effective by a great number of experiments mentioned later in the paper.In the discussion of image fusion methods based on spatial domain, the paper introduces the simple image fusion method and then suggests, on the basis of fusion with weighted average, to figure out the weighted average coefficient by using PCA transform, which helps to form images of better fusion effect. And then describes the domainal properties-based method of which the overall performance is superior to that of simple fusion methods.The fourth chapter of the paper focuses on the wavelet-transform-based image fusion method of which the basic theories, multi-resolution analysis and Mallat fast arithmetic are discussed. Then the level-based and region-based fusion rules and implement of these are discussed and a novel fusion rule is raised. Through a series of experiments, what can be done to fusion performance by wavelet radix, the number of decomposition levels, different fusion of rules, the size of region for measure or threshold is studied according to valuation methods proposed in the paper.At the end of the paper, the paper brings forward a new fusion method on the basis of image segments. Experiments show that it can sustain useful information, realize the complementation of information and therefore have such advantages as simple arithmetic, fast runtime and better effect. Image fusion technology can be applied in many fields, and its application in medicine is also explored, which further support the methods advanced in the paper.更多还原