Digital Image Processing, 4/e (GE-Paperback) (數位影像處理（第4版）)

《影像處理與電腦視覺》是一本適用於影像處理和電腦視覺課程的教材。

這本書已經成為數位影像處理研究的基礎教材已經有40年的歷史。適合大學高年級和研究生一年級的學生，需要具備數學分析、向量、矩陣、概率、統計、線性系統和電腦編程的背景知識。與之前的版本一樣，本版的重點仍然是基礎知識。

這本書的第四版是為了慶祝其40週年而出版的，基於對來自30個國家的150所學院的教師、學生和獨立讀者的廣泛調查。他們的反饋導致了對深度學習和深度神經網絡等主題的擴展或新增覆蓋範圍，包括卷積神經網絡、尺度不變特徵變換（SIFT）、極值穩定區域（MSERs）、圖割、k-means聚類和超像素、主動輪廓（snakes和level sets）以及精確的直方圖匹配。在將影像變換的材料重新組織成更具連貫性的演示和空間核和空間濾波的討論方面進行了重大改進。在整本書中，對例子和作業練習進行了重大修訂和增加。首次，我們在每章結尾增加了MATLAB項目，並編制了支持學生和教師的輔助材料，包括解答、影像數據庫和示例代碼。

1 Introduction

1.1 What is Digital Image Processing?

1.2 The Origins of Digital Image Processing

1.3 Examples of Fields that Use Digital Image Processing

   Gamma-Ray Imaging

   X-Ray Imaging

   Imaging in the Ultraviolet Band

   Imaging in the Visible and Infrared Bands

   Imaging in the Microwave Band

   Imaging in the Radio Band

   Other Imaging Modalities

1.4 Fundamental Steps in Digital Image Processing

1.5 Components of an Image Processing System

 

2 Digital Image Fundamentals

2.1 Elements of Visual Perception

   Structure of the Human Eye

   Image Formation in the Eye

   Brightness Adaptation and Discrimination

2.2 Light and the Electromagnetic Spectrum

2.3 Image Sensing and Acquisition

   Image Acquisition Using a Single Sensing Element

   Image Acquisition Using Sensor Strips

   Image Acquisition Using Sensor Arrays

   A Simple Image Formation Model

2.4 Image Sampling and Quantization

   Basic Concepts in Sampling and Quantization

   Representing Digital Images

   Linear vs. Coordinate Indexing

   Spatial and Intensity Resolution

   Image Interpolation

2.5 Some Basic Relationships Between Pixels

   Neighbors of a Pixel

   Adjacency, Connectivity, Regions, and Boundaries

   Distance Measures

2.6 Introduction to the Basic Mathematical Tools Used in Digital Image Processing

   Elementwise versus Matrix Operations

   Linear versus Nonlinear Operations

   Arithmetic Operations

   Set and Logical Operations

Basic Set Operations

Logical Operations

Fuzzy Sets

   Spatial Operations

Single-Pixel Operations

   Neighborhood Operations

   Geometric Transformations

   Image Registration

   Vector and Matrix Operations

   Image Transforms

   Probability and Random Variables

 

3 Intensity Transformations and Spatial Filtering

3.1 Background

   The Basics of Intensity Transformations and Spatial Filtering

   About the Examples in this Chapter

3.2 Some Basic Intensity Transformation Functions

   Image Negatives

   Log Transformations

   Power-Law (Gamma) Transformations

   Piecewise Linear Transformation Functions

Contrast Stretching

   Intensity-Level Slicing

   Bit-Plane Slicing

3.3 Histogram Processing

   Histogram Equalization

   Histogram Matching (Specification)

   Exact Histogram Matching (Specification)

   Foundation

   Ordering

                        Computing the neighborhood averages and extracting the K-tuples:

   Exact Histogram Specification Algorithm

   Local Histogram Processing

   Using Histogram Statistics for Image Enhancement

3.4 Fundamentals of Spatial Filtering

   The Mechanics of Linear Spatial Filtering

   Spatial Correlation and Convolution

   Separable Filter Kernels

   Some Important Comparisons Between Filtering in the Spatial and Frequency Domains

   A Word about how Spatial Filter Kernels are Constructed

3.5 Smoothing (Lowpass) Spatial Filters

   Box Filter Kernels

   Lowpass Gaussian Filter Kernels

   Order-Statistic (Nonlinear) Filters

3.6 Sharpening (Highpass) Spatial Filters

   Foundation

   Using the Second Derivative for Image Sharpening—The Laplacian

   Unsharp Masking and Highboost Filtering

   Using First-Order Derivatives for Image Sharpening—The Gradient

3.7 Highpass, Bandreject, and Bandpass Filters from Lowpass Filters

3.8 Combining Spatial Enhancement Methods

3.9 Using Fuzzy Techniques for Intensity Transformations and Spatial Filtering

   Introduction

   Principles of Fuzzy Set Theory

   Definitions

   Some Common Membership Functions

   Using Fuzzy Sets

   Using Fuzzy Sets for Intensity Transformations

   Using Fuzzy Sets for Spatial Filtering

 

4 Filtering in the Frequency Domain

4.1 Background

   A Brief History of the Fourier Series and Transform

   About the Examples in this Chapter

4.2 Preliminary Concepts

   Complex Numbers

   Fourier Series

   Impulses and their Sifting Properties

   The Fourier Transform of Functions of One Continuous Variable

   Convolution

4.3 Sampling and the Fourier Transform of Sampled Functions

   Sampling

   The Fourier Transform of Sampled Functions

   The Sampling Theorem

   Aliasing

   Function Reconstruction (Recovery) from Sampled Data

4.4 The Discrete Fourier Transform of One Variable

   Obtaining the DFT from the Continuous Transform of a Sampled Function

   Relationship Between the Sampling and Frequency Intervals

4.5 Extensions to Functions of Two Variables

   The 2-D Impulse and Its Sifting Property

   The 2-D Continuous Fourier Transform Pair

   2-D Sampling and the 2-D Sampling Theorem

   Aliasing in Images

   Extensions from 1-D Aliasing

   Image Resampling and Interpolation

   Aliasing and Moiré Patterns

   The 2-D Discrete Fourier Transform and Its Inverse

4.6 Some Properties of the 2-D DFT and IDFT

   Relationships Between Spatial and Frequency Intervals

   Translation and Rotation

   Periodicity

   Symmetry Properties

   Fourier Spectrum and Phase Angle

   The 2-D Discrete Convolution Theorem

   Summary of 2-D Discrete Fourier Transform Properties

4.7 The Basics of Filtering in the Frequency Domain

   Additional Characteristics of the Frequency Domain

   Frequency Domain Filtering Fundamentals

   Summary of Steps for Filtering in the Frequency Domain

   Correspondence Between Filtering in the Spatial and

   Frequency Domains

4.8 Image Smoothing Using Lowpass Frequency Domain Filters

   Ideal Lowpass Filters

   Gaussian Lowpass Filters

   Butterworth Lowpass Filters

   Additional Examples of Lowpass Filtering

4.9 Image Sharpening Using Highpass Filters

   Ideal, Gaussian, and Butterworth Highpass Filters from Lowpass Filters

   The Laplacian in the Frequency Domain

   Unsharp Masking, High-boost Filtering, and High-Frequency-Emphasis Filtering

   Homomorphic Filtering

4.10 Selective Filtering

   Bandreject and Bandpass Filters

   Notch Filters

4.11 The Fast Fourier Transform

   Separability of the 2-D DFT

   Computing the IDFT Using a DFT Algorithm

   The Fast Fourier Transform (FFT)

 

5 Image Restoration and Reconstruction

5.1 A Model of the Image Degradation/Restoration Process

5.2 Noise Models

   Spatial and Frequency Properties of Noise

   Some Important Noise Probability Density Functions

   Gaussian Noise

   Rayleigh Noise

   Erlang (Gamma) Noise

   Exponential Noise

   Uniform Noise

   Salt-and-Pepper Noise

   Periodic Noise

   Estimating Noise Parameters

5.3 Restoration in the Presence of Noise Only—Spatial Filtering

   Mean Filters

   Arithmetic Mean Filter

   Geometric Mean Filter

   Harmonic Mean Filter

   Contraharmonic Mean Filter

   Order-Statistic Filters

   Median Filter

   Max and Min Filters

   Midpoint Filter

   Alpha-Trimmed Mean Filter

   Adaptive Filters

   Adaptive, Local Noise Reduction Filter

   Adaptive Median Filter

5.4 Periodic Noise Reduction Using Frequency Domain Filtering

   More on Notch Filtering

   Optimum Notch Filtering

5.5 Linear, Position-Invariant Degradations

5.6 Estimating the Degradation Function

   Estimation by Image Observation

   Estimation by Experimentation

   Estimation by Modeling

5.7 Inverse Filtering

5.8 Minimum Mean Square Error (Wiener) Filtering

5.9 Constrained Least Squares Filtering

5.10 Geometric Mean Filter

5.11 Image Reconstruction from Projections

   Introduction

   Principles of X-ray Computed Tomography (CT)

   Projections and the Radon Transform

   Backprojections

   The Fourier-Slice Theorem

   Reconstruction Using Parallel-Beam Filtered Backprojections

   Reconstruction Using Fan-Beam Filtered Backprojections

 

6 Wavelet and Other Image Transforms

6.1 Preliminaries

6.2 Matrix-based Transforms

   Rectangular Arrays

   Complex Orthonormal Basis Vectors

   Biorthonormal Basis Vectors

6.3 Correlation

6.4 Basis Functions in the Time-Frequency Plane

6.5 Basis Images

6.6 Fourier-Related Transforms

   The Discrete hartley Transform

   The Discrete Cosine Transform

   The Discrete Sine Transform

6.7 Walsh-Hadamard Transforms

6.8 Slant Transform

6.9 Haar Transform

6.10 Wavelet Transforms

   Scaling Functions

   Wavelet Functions

   Wavelet Series Expansion

   Discrete Wavelet Transform in One Dimension

   The Fast Wavelet Transform

   Wavelet Transforms in Two Dimensions

   Wavelet Packets

 

7 Color Image Processing

7.1 Color Fundamentals

7.2 Color Models

   The RGB Color Model

   The CMY and CMYK Color Models

   The HSI Color Model

   Converting Colors from RGB to HSI

   Converting Colors from HSI to RGB

   Manipulating HSI Component Images

   A Device Independent Color Model

7.3 Pseudocolor Image Processing

   Intensity Slicing and Color Coding

   Intensity to Color Transformations

7.4 Basics of Full-Color Image Processing

7.5 Color Transformations

   Formulation

   Color Complements

   Color Slicing

   Tone and Color Corrections

   Histogram Processing of Color Images

7.6 Color Image Smoothing and Sharpening

   Color Image Smoothing

   Color Image Sharpening

7.7 Using Color in Image Segmentation

   Segmentation in HSI Color Space

   Segmentation in RGB Space

   Color Edge Detection

7.8 Noise in Color Images

7.9 Color Image Compression

 

8 Image Compression and Watermarking

8.1 Fundamentals

   Coding Redundancy

   Spatial and Temporal Redundancy

   Irrelevant Information

   Measuring Image Information

   Shannon’s First Theorem

   Fidelity Criteria

   Image Compression Models

   The Encoding or Compression Process

   The Decoding or Decompression Process

   Image Formats, Containers, and Compression Standards

8.2 Huffman Coding

8.3 Golomb Coding

8.4 Arithmetic Coding

   Adaptive context dependent probability estimates

8.5 LZW Coding

8.6 Run-length Coding

   One-dimensional CCITT compression

   Two-dimensional CCITT compression

8.7 Symbol-based Coding

   JBIG2 compression

8.8 Bit-plane Coding

8.9 Block Transform Coding

   Transform selection

   Subimage size selection

   Bit allocation

   Zonal Coding Implementation

   Threshold Coding Implementation

   JPEG

8.10 Predictive Coding

   Lossless predictive coding

   Motion compensated prediction residuals

   Lossy predictive coding

   Optimal predictors

   Optimal quantization

8.11 Wavelet Coding

   Wavlet selection

   Decomposition level selection

   Quantizer design

   JPEG-2000

8.12 Digital Image Watermarking

 

9 Morphological Image Processing

9.1 Preliminaries

9.2 Erosion and Dilation

   Erosion

   Dilation

   Duality

9.3 Opening and Closing

9.4 The Hit-or-Miss Transform

9.5 Some Basic Morphological Algorithms

   Boundary Extraction

   Hole Filling

   Extraction of Connected Components

   Convex Hull

   Thinning

   Thickening

   Skeletons

   Pruning

9.6 Morphological Reconstruction

   Geodesic Dilation and Erosion

   Morphological Reconstruction by Dilation and by Erosion

   Sample Applications

   Opening by Reconstruction

   Automatic Algorithm for Filling Holes

   Border Clearing

9.7 Summary of Morphological Operations on Binary Images

9.8 Grayscale Morphology

   Grayscale Erosion and Dilation

   Grayscale Opening and Closing

   Some Basic Grayscale Morphological Algorithms

   Morphological Smoothing

   Morphological Gradient

   Top-Hat and Bottom-Hat Transformations

   Granulometry

   Textural Segmentation

   Grayscale Morphological Reconstruction

 

10 Image Segmentation I: Edge Detection,

   Thresholding, and Region Detection

10.1 Fundamentals

10.2 Point, Line, and Edge Detection

   Background

   Detection of Isolated Points

   Line Detection

   Edge Models

   Basic Edge Detection

   The Image Gradient and Its Properties

   Gradient Operators

   Combining the Gradient with Thresholding

   More Advanced Techniques for Edge Detection

   The Marr-Hildreth Edge Detector

   The Canny Edge Detector

   Linking Edge Points

   Local Processing

   Global Processing Using the Hough Transform

10.3 Thresholding

   Foundation

   The Basics of Intensity Thresholding

   The Role of Noise in Image Thresholding

   The Role of Illumination and Reflectance in Image Thresholding

   Basic Global Thresholding

   Optimum Global Thresholding Using Otsu’s Method

   Using Image Smoothing to Improve Global Thresholding

   Using Edges to Improve Global Thresholding

   Multiple Thresholds

   Variable Thresholding

   Variable Thresholding Based on Local Image Properties

   Variable Thresholding Based on Moving Averages

10.4 Segmentation by Region Growing and by Region Splitting and Merging

   Region Growing

   Region Splitting and Merging

10.5 Region Segmentation Using Clustering and Superpixels

   Region Segmentation using K-Means Clustering

   Region Segmentation using Superpixels

   SLIC Superpixel Algorithm

   Specifying the Distance Measure

10.6 Region Segmentation Using Graph Cuts

   Images as Graphs

   Minimum Graph Cuts

   Computing Minimal Graph Cuts

   Graph Cut Segmentation Algorithm

10.7 Segmentation Using Morphological Watersheds

   Background

   Dam Construction

   Watershed Segmentation Algorithm

   The Use of Markers

10.8 The Use of Motion in Segmentation

   Spatial Techniques

   A Basic Approach

   Accumulative Differences

   Establishing a Reference Image

   Frequency Domain Techniques

 

11 Image Segmentation II: Active Contours: Snakes and Level Sets

11.1 Background

11.2 Image Segmentation Using Snakes

   Explicit (Parametric) Representation of Active Contours

   Derivation of the Fundamental Snake Equation

   Iterative Solution of the Snake Equation

   External Force Based on the Magnitude of the Image

   Gradient (MOG)

   External Force Based on Gradient Vector Flow (GVF)

11.3 Segmentation Using Level Sets

   Implicit Representation of Active Contours

   Derivation of the Level Set Equation

   Discrete (Iterative) Solution of The Level Set Equation

   Curvature

   Specifying, Initializing, and Reinitializing Level Set Functions

   Force Functions Based Only on Image Properties

   Edge/Curvature-Based Forces

   Region/Curvature-Based Forces

   Improving the Computational Performance of Level Set Algorithms

 

12 Feature Extraction

12.1 Background

12.2 Boundary Preprocessing

   Boundary Following (Tracing)

   Chain Codes

   Freeman Chain Codes

   Slope Chain Codes

   Boundary Approximations Using Minimum-Perimeter Polygons

   Foundation

   MPP Algorithm

   Signatures

   Skeletons, Medial Axes, and Distance Transforms

12.3 Boundary Feature Descriptors

   Some Basic Boundary Descriptors

   Shape Numbers

   Fourier Descriptors

   Statistical Moments

12.4 Region Feature Descriptors

   Some Basic Descriptors

   Topological Descriptors

   Texture

   Statistical Approaches

   Spectral Approaches

   Moment Invariants

12.5 Principal Components as Feature Descriptors

12.6 Whole-Image Features

   The Harris-Stephens Corner Detector

   Maximally Stable Extremal Regions (MSERs)

12.7 Scale-Invariant Feature Transform (SIFT)

   Scale Space

   Detecting Local Extrema

   Finding the Initial Keypoints

   Improving the Accuracy of Keypoint Locations

   Eliminating Edge Responses

   Keypoint Orientation

   Keypoint Descriptors

   Summary of the SIFT Algorithm

 

13 Image Pattern Classification

13.1 Background

13.2 Patterns and Pattern Classes

   Pattern Vectors

   Structural Patterns

13.3 Pattern Classification by Prototype Matching

   Minimum-Distance Classifier

   Using Correlation for 2-D prototype matching

   Matching SIFT Features

   Matching Structural Prototypes

   Matching Shape Numbers

   String Matching

13.4 Optimum (Bayes) Statistical Classifiers

   Derivation of the Bayes Classifier

   Bayes Classifier for Gaussian Pattern Classes

13.5 Neural Networks and Deep Learning

   Background

   The Perceptron

   Multilayer Feedforward Neural Networks

   Model of an Artificial Neuron

   Interconnecting Neurons to Form a Fully Connected Neural Network

   Forward Pass Through a Feedforward Neural Network

   The Equations of a Forward Pass

   Matrix Formulation

   Using Backpropagation to Train Deep Neural Networks

   The Equations of Backpropagation

   Matrix Formulation

13.6 Deep Convolutional Neural Networks

   A Basic CNN Architecture

   Basics of How a CNN Operates

   Neural Computations in a CNN

   Multiple Input Images

   The Equations of a Forward Pass Through a CNN

   The Equations of Backpropagation Used to Train CNNs

13.7 Some Additional Details of Implementation

   Bibliography

   Index

1 簡介

1.1 什麼是數位影像處理？

1.2 數位影像處理的起源

1.3 使用數位影像處理的領域範例

   伽瑪射線成像

   X射線成像

   紫外線成像

   可見光和紅外線成像

   微波成像

   無線電成像

   其他成像模式

1.4 數位影像處理的基本步驟

1.5 影像處理系統的組成部分

 

2 數位影像基礎

2.1 視覺感知的元素

   人眼的結構

   眼睛中的影像形成

   亮度適應和區分度

2.2 光和電磁譜

2.3 影像感測和擷取

   使用單一感測元素的影像擷取

   使用感測條帶的影像擷取

   使用感測陣列的影像擷取

   簡單的影像形成模型

2.4 影像取樣和量化

   取樣和量化的基本概念

   數位影像的表示

   線性與座標索引

   空間和強度解析度

   影像插值

2.5 像素之間的一些基本關係

   像素的鄰居

   鄰接性、連通性、區域和邊界

   距離度量

2.6 數位影像處理中使用的基本數學工具介紹

   逐元素與矩陣操作

   線性與非線性操作

   算術操作

   集合和邏輯操作

基本集合操作

邏輯操作

模糊集合

   空間操作

單像素操作

   鄰域操作

   幾何轉換

   影像配準

   向量和矩陣操作

   影像轉換

   概率和隨機變數

 

3 強度轉換和空間濾波

3.1 背景

   強度轉換和空間濾波的基礎知識

   本章節中的範例說明

3.2 一些基本的強度轉換函數

   影像的負片

   對數轉換

   冪律（伽瑪）轉換

   分段線性轉換函數

對比度拉伸

   強度級別切割

   位平面切割

3.3 直方圖處理

   直方圖均衡化

   直方圖匹配（規格化）

   精確直方圖匹配（規格化）

   基礎

   排序

                        計算鄰域平均值並提取K元組：

   精確直方圖規格化演算法

   局部直方圖處理

   使用直方圖統計進行影像增強

3.4 空間濾波的基礎

   線性空間濾波的機制

   空間相關和卷積