图像融合技术：理论、技巧与应用

需积分: 20 51 浏览量更新于2024-07-22 1 收藏 2.78MB PDF 举报

"Image Fusion"是一本关于图像融合的权威教材，由H.B. Mitchell撰写，由ABC出版社出版。这本书深入探讨了图像融合的理论、技术和应用，是该领域的经典著作，可供读者免费阅读。图像融合是计算机视觉和图像处理领域的一个重要课题，它涉及到多源图像的整合，以生成具有更丰富信息或更高解析度的新图像。在这个过程中，不同的图像源可能包括不同传感器（如可见光、红外、超声波等）获取的数据，或者同一场景在不同时间、不同条件下的图像。书中的内容可能涵盖了以下几个关键知识点： 1. **融合理论**：首先，书中会介绍图像融合的基本概念，包括为什么需要图像融合，以及融合在哪些应用中具有重要意义。它可能会讲解信息融合理论，比如熵理论、信息论等，这些理论是理解图像融合基础的基石。 2. **融合技术**：书中会详细介绍各种图像融合技术，例如像素级融合、特征级融合和决策级融合。像素级融合通常涉及对图像像素值的简单操作，如平均或加权平均；特征级融合则考虑图像的高级特征，如边缘、纹理、颜色等；决策级融合在更高的抽象层面上进行，可能涉及机器学习算法来决定最佳组合。 3. **应用领域**：H.B. Mitchell可能会讨论图像融合在多个领域的应用，如医学成像（用于提高诊断精度）、遥感（用于环境监测和灾害评估）、军事侦察（用于目标识别和态势感知）、机器人导航（用于增强环境理解）等。 4. **算法实现**：书中可能包含一些具体的图像融合算法，如PCA（主成分分析）、DWT（离散小波变换）、MST（最小生成树）、FSF（模糊集融合）等，并提供实例分析和代码示例，帮助读者理解和实践。 5. **评估与挑战**：作者可能会讨论如何评估融合结果的质量，以及在实际应用中面临的挑战，如计算复杂性、实时性需求、数据隐私和版权问题等。 6. **法律与版权**：提及的版权信息表明，虽然内容可以被复制，但必须遵循德国版权法的规定，未经许可，不得非法使用或复制该作品的全部或部分内容。这本教材对想要深入研究图像融合的学者、研究人员以及从业人员来说是一份宝贵的资源，它不仅提供了理论知识，还可能包含实践指导，有助于读者掌握这一关键技术。

Chapter 1

Introduction

Abstract. The subject of this book is image fusion which we deﬁne as the process

of combining multiple input images into a single composite image. Our aim is to

create from the collection of input images a single output image which contains a

better description of the scene than the one provided by any of the individual in-

put images. The output image should therefore be more useful for human visual

perception or for machine perception. The basic problem of image fusion is one of

determining the best procedure for combining the multiple input images. The view

adopted in this book is that combining multiple images with aprioriinformation

is best handled within a statistical framework. In particular we shall restrict our-

selves to classical and robust statistical approaches, Bayesian methods, sub-space

and wavelet techniques.

1.1 Synergy

The principal motivation for image fusion is to improve the quality of the infor-

mation contained in the output image in a process known as synergy. A study of

existing image fusion techniques and applications shows that image fusion can pro-

vide us with an output image with an improved quality. In this case, the beneﬁts of

image fusion include:

1. Extended range of operation.

2. Extended spatial and temporal coverage.

3. Reduced uncertainty.

4. Increased reliability.

5. Robust system performance.

6. Compact representation of information.

Traditionally, the input images are captured by the same camera at different times

or are captured by different cameras at the same time. However, in the deﬁnition of

image fusion we shall also include the case when the input images are derived from

the same “base” image but which have undergone different processing algorithms.

2 1 Introduction

The following examples illustrate the image fusion synergy process. The ﬁrst

example deals with input images which are captured by the same camera at different

times. The next example deals with input images captured by different cameras at

the same time, while the third example deals with input images which are derived

from the same base image and which are processed differently.

Example 1.1. Multiple Camera Surveillance Systems [9]. The increasing de-

mand for security by society has led to a growing need for surveillance ac-

tivities in many environments. For example, the surveillance of a wide-area

urban site may be provided by periodically scanning the area with a single

narrow ﬁeld-of-view camera. The temporal coverage is, however, limited by

the time required for the camera to execute one scan. By using multiple cam-

eras we reduce the mean time between scans and thereby increase the temporal

coverage.

Example 1.2. Multispectral Bilateral Video Fusion [2]. A signiﬁcant problem

in night vision imagery is that while an infra-red (IR) image provides a bright

and relatively low-noise view of a dark environment, it can be difﬁcult to in-

terpret due to inconsistences with the corresponding visible-spectrum image.

In bilateral fusion we enhance a visible video input using information from

a spatially and temporally registered IR video input. Our goal is to create a

video that appears as if it was imaged only in the visible spectrum and under

more ideal exposure conditions than actually existed.

Example 1.3. Color Image Segmentation [11]. A signiﬁcant problem in com-

puter vision is the reliable segmentation a base image into meaningful labeled

segments. In ensemble image segmentation we generate an ensemble of color

input images by transforming the base image in different ways. Each input im-

age is separately segmented using a simple segmentation algorithm. By fusing

the multiple segmented images we are able to substantially improve both the

accuracy and the reliability of the segmentation process.

1.2 Image Fusion Process

Fig. 1.1 shows the principal processes in a generic image fusion processing chain

for the case when the output is a single fused image



I. The principal processes in

the chain are:

4 1 Introduction

1.3 Common Representational Block

The principal function in the common representational format block are:

Spatial Alignment. The input images are spatially aligned into the same geo-

metric base. Without a common geometric base any information derived from a

given input image cannot be associated with other spatial information. The accu-

rate spatial alignment of the input images is therefore a necessary condition for

image fusion. Note. After spatial alignment the input images are re-sampled and

if necessary the gray-levels of the input images are interpolated.

Temporal Alignment. The spatially aligned input images are temporally aligned

to a common time. This step is only required if the input images are changing or

evolving in time. In this case the accurate temporal alignment of the input images

is a necessary condition for image fusion.

Feature Extraction. Characteristic features are extracted from the spatially and

temporally aligned input images. The output is one or more feature maps for each

input image

[1]

Decision labeling. Pixels in each spatially and temporally aligned input image

or feature map are labeled according to a given criteria. The output is a set of

decision maps.

Semantic Equivalence. In order for the input images, feature maps or decision

maps to be fused together they must refer to the same object or phenomena. The

process of causally linking the different inputs to a common object or phenomena

is known as semantic equivalence.

Radiometric Calibration. The spatially, temporally and semantically aligned in-

put images and feature maps are converted to a common measurement scale. This

process is known as radiometric calibration.

When the input is a set of K multiple image sequences I

(t),t ∈ [T

],k ∈

{1,2,...,K}, the output is a fused image sequence



I(t),t ∈ [T

], feature map

sequence



F(t) or decision map sequence



D(t). In this case we replace the spatial

alignment and temporal alignment blocks in Fig. 1.1 with a single spatial-temporal

alignment block which performs both functions simultaneously.

Figs. 1.2–1.4 shows the adaption of the generic processing chain to Ex 1.1–1.3.

1.4 Image Fusion Block

In the image fusion block we fuse together the information contained in the multiple

input images after conversion into a common representational format. The common

representational format may take the form of an image I, a feature map F or a

decision map D. Very often we shall not differentiate between I, F and D and in this

case we shall refer to all three as an “image”.

We ﬁnd it convenient to further divide the fusion algorithms into arithmetic, sub-

space and multi-scale techniques [5].

A feature is any distinguishing property or attribute of an image. Examples of features

used in image fusion are: edges, lines, patterns and color.

剩余247页未读，继续阅读

qq_27500949

粉丝: 0
资源: 4

图像融合技术：理论、技巧与应用

image-fusion

fusionimage

image fusion tools

ImageFusion

imagefusion_hsi_exceptbi8_imagefusion_imagefusion_图像变换_

imagefusion.rar_image fusion_imagefusion_sumkqh_全色 多光谱_金字塔融合

image-fusion.rar_PCA融合_fusion_ihs_image fusion_pca image fusion

Multi-Focus Image Fusion Based on Spatial.rar_image fusion_image

image-fusion-master_image_imagefusion_

NSCT-SF-PCNN-ImageFusion-Toolbox_imagefusion_

最新资源

imagefusion.rar_image fusion_imagefusion_sumkqh_全色多光谱_金字塔融合