Details

<p>Contributors xv</p> <p>Acknowledgments xvii</p> <p>About the Companion Website xix</p> <p><b>1 Introduction 1</b></p> <p>1.1 Organization of the Book 3</p> <p>1.2 Further Reading 5</p> <p>References 5</p> <p><b>2 Image Orientation Detection 9</b></p> <p>2.1 Introduction 9</p> <p>2.2 Geometric Gradient and Geometric Smoothing 13</p> <p>2.2.1 Calculating Geometric Gradients 15</p> <p>2.3 Comparison of Geometric Gradients and Intensity Gradients 18</p> <p>2.4 Finding the Rotational Difference between Two Images 21</p> <p>2.5 Performance Evaluation 23</p> <p>2.5.1 Reliability 23</p> <p>2.5.2 Accuracy 31</p> <p>2.5.3 Computational Complexity 32</p> <p>2.6 Registering Images with a Known Rotational Difference 34</p> <p>2.7 Discussion 36</p> <p>2.8 Further Reading 37</p> <p>References 40</p> <p><b>3 Feature Point Detection 43</b></p> <p>3.1 Introduction 43</p> <p>3.2 Variant Features 44</p> <p>3.2.1 Central Moments 44</p> <p>3.2.2 Uniqueness 48</p> <p>3.3 Invariant Features 50</p> <p>3.3.1 Rotation-Invariant Features 50</p> <p>3.3.1.1 Laplacian of Gaussian (LoG) Detector 51</p> <p>3.3.1.2 Entropy 53</p> <p>3.3.1.3 InvariantMoments 55</p> <p>3.3.2 SIFT: A Scale-and Rotation-Invariant Point Detector 58</p> <p>3.3.3 Radiometric-Invariant Features 60</p> <p>3.3.3.1 Harris Corner Detector 60</p> <p>3.3.3.2 Hessian Corner Detector 63</p> <p>3.4 Performance Evaluation 64</p> <p>3.5 Further Reading 68</p> <p>References 68</p> <p><b>4 FeatureLineDetection 75</b></p> <p>4.1 Hough Transform Using Polar Equation of Lines 79</p> <p>4.2 Hough Transform Using Slope and y-Intercept Equation of Lines 82</p> <p>4.3 Line Detection Using Parametric Equation of Lines 86</p> <p>4.4 Line Detection by Clustering 89</p> <p>4.5 Line Detection by Contour Tracing 92</p> <p>4.6 Line Detection by Curve Fitting 95</p> <p>4.7 Line Detection by Region Subdivision 101</p> <p>4.8 Comparison of the Line Detection Algorithms 106</p> <p>4.8.1 Sensitivity to Noise 106</p> <p>4.8.2 Positional and Directional Errors 106</p> <p>4.8.3 Length Accuracy 109</p> <p>4.8.4 Speed 109</p> <p>4.8.5 Quality of Detected Lines 109</p> <p>4.9 Revisiting Image Dominant Orientation Detection 117</p> <p>4.10 Further Reading 121</p> <p>References 125</p> <p><b>5 Finding Homologous Points 133</b></p> <p>5.1 Introduction 133</p> <p>5.2 Point Pattern Matching 134</p> <p>5.2.1 Parameter Estimation by Clustering 137</p> <p>5.2.2 Parameter Estimation by RANSAC 141</p> <p>5.3 Point Descriptors 146</p> <p>5.3.1 Histogram-Based Descriptors 147</p> <p>5.3.2 SIFT Descriptor 148</p> <p>5.3.3 GLOH Descriptor 151</p> <p>5.3.4 Composite Descriptors 152</p> <p>5.3.4.1 Hu InvariantMoments 152</p> <p>5.3.4.2 Complex Moments 152</p> <p>5.3.4.3 Cornerness Measures 153</p> <p>5.3.4.4 Power Spectrum Features 154</p> <p>5.3.4.5 Differential Features 155</p> <p>5.3.4.6 Spatial Domain Features 155</p> <p>5.4 SimilarityMeasures 160</p> <p>5.4.1 Correlation Coefficient 160</p> <p>5.4.2 Minimum Ratio 161</p> <p>5.4.3 Spearman’s ;; 161</p> <p>5.4.4 Ordinal Measure 162</p> <p>5.4.5 Correlation Ratio 162</p> <p>5.4.6 Shannon Mutual Information 164</p> <p>5.4.7 Tsallis Mutual Information 165</p> <p>5.4.8 F-Information 166</p> <p>5.5 Distance Measures 167</p> <p>5.5.1 Sum of Absolute Differences 167</p> <p>5.5.2 Median of Absolute Differences 167</p> <p>5.5.3 Square Euclidean Distance 168</p> <p>5.5.4 Intensity-Ratio Variance 168</p> <p>5.5.5 Rank Distance 169</p> <p>5.5.6 Shannon Joint Entropy 169</p> <p>5.5.7 Exclusive F-Information 170</p> <p>5.6 TemplateMatching 170</p> <p>5.6.1 Coarse-to-Fine Matching 171</p> <p>5.6.2 MultistageMatching 172</p> <p>5.6.3 Rotationally InvariantMatching 173</p> <p>5.6.4 Gaussian-Weighted TemplateMatching 174</p> <p>5.6.5 Template Matching in Different Modality Rotated Images 175</p> <p>5.7 Robust Parameter Estimation 178</p> <p>5.7.1 Ordinary Least-Squares Estimator 180</p> <p>5.7.2 Weighted Least-Squares Estimator 182</p> <p>5.7.3 Least Median of Squares Estimator 184</p> <p>5.7.4 Least Trimmed Squares Estimator 184</p> <p>5.7.5 Rank Estimator 185</p> <p>5.8 Finding Optimal Transformation Parameters 193</p> <p>5.9 Performance Evaluation 193</p> <p>5.10 Further Reading 197</p> <p>References 200</p> <p><b>6 Finding Homologous Lines 215</b></p> <p>6.1 Introduction 215</p> <p>6.2 Determining Transformation Parameters from Line Parameters 215</p> <p>6.3 Finding Homologous Lines by Clustering 221</p> <p>6.3.1 Finding the Rotation Parameter 222</p> <p>6.3.2 Finding the Translation Parameters 223</p> <p>6.4 Finding Homologous Lines by RANSAC 229</p> <p>6.5 Line Grouping Using Local Image Information 232</p> <p>6.6 Line Grouping Using Vanishing Points 235</p> <p>6.6.1 Methods Searching the Image Space 235</p> <p>6.6.2 Methods Searching the Polar Space 236</p> <p>6.6.3 Methods Searching the Gaussian Sphere 236</p> <p>6.6.4 A Method Searching Both Image and Gaussian Sphere 237</p> <p>6.6.5 Measuring the Accuracy of Detected Vanishing Points 244</p> <p>6.6.6 Discussion 247</p> <p>6.7 Robust Parameter Estimation Using Homologous Lines 253</p> <p>6.8 Revisiting Image Dominant Orientation Detection 255</p> <p>6.9 Further Reading 256</p> <p>References 257</p> <p><b>7 Nonrigid Image Registration 261</b></p> <p>7.1 Introduction 261</p> <p>7.2 Finding Homologous Points 262</p> <p>7.2.1 Coarse-to-Fine Matching 262</p> <p>7.2.2 Correspondence by Template Matching 269</p> <p>7.3 Outlier Removal 274</p> <p>7.4 Elastic Transformation Models 278</p> <p>7.4.1 Surface Spline (SS) Interpolation 280</p> <p>7.4.2 Piecewise Linear (PWL) Interpolation 282</p> <p>7.4.3 Moving Least Squares (MLS) Approximation 283</p> <p>7.4.4 Weighted Linear (WL) Approximation 285</p> <p>7.4.5 Performance Evaluation 287</p> <p>7.4.6 Choosing the Right Transformation Model 291</p> <p>7.5 Further Reading 292</p> <p>References 293</p> <p><b>8 Volume Image Registration 299</b></p> <p>8.1 Introduction 299</p> <p>8.2 Feature Point Detection 301</p> <p>8.2.1 Central Moments 301</p> <p>8.2.2 Entropy 302</p> <p>8.2.3 LoG Operator 302</p> <p>8.2.4 First-Derivative Intensities 303</p> <p>8.2.5 Second-Derivative Intensities 304</p> <p>8.2.6 Speed-Up Considerations in Feature Point Detection 305</p> <p>8.2.7 Evaluation of Feature Point Detectors 305</p> <p>8.3 Finding Homologous Points 307</p> <p>8.3.1 Finding Initial Homologous Points Using Image Descriptors 310</p> <p>8.3.2 Finding Initial Homologous Points by Template Matching 313</p> <p>8.3.3 Finding Final Homologous Points from Coarse to Fine 315</p> <p>8.3.4 Finding the Final Homologous Points by Outlier Removal 320</p> <p>8.4 Transformation Models for Volume Image Registration 321</p> <p>8.4.1 Volume Spline 323</p> <p>8.4.2 Weighted Rigid Transformation 325</p> <p>8.4.3 Computing the Overall Transformation 327</p> <p>8.5 Performance Evaluation 330</p> <p>8.5.1 Accuracy 330</p> <p>8.5.2 Reliability 333</p> <p>8.5.3 Speed 333</p> <p>8.6 Further Reading 335</p> <p>References 337</p> <p><b>9 Validation Methods 343</b></p> <p>9.1 Introduction 343</p> <p>9.2 Validation Using Simulation Data 344</p> <p>9.3 Validation Using a Gold Standard 345</p> <p>9.4 Validation by an Expert Observer 347</p> <p>9.5 Validation Using a Consistency Measure 348</p> <p>9.6 Validation Using a Similarity/DistanceMeasure 350</p> <p>9.7 Further Reading 351</p> <p>References 352</p> <p><b>10 Video Image Registration 357</b></p> <p>EdgardoMolina,Wai Lun Khoo, Hao Tang, and Zhigang Zhu</p> <p>10.1 Introduction 357</p> <p>10.2 Motion Modeling 358</p> <p>10.2.1 The Motion Field of Rigid Objects 358</p> <p>10.2.2 Motion Models 360</p> <p>10.2.2.1 Pure Rotation and a 3-D Scene 361</p> <p>10.2.2.2 General Motion and a Planar Scene 362</p> <p>10.2.2.3 TranslationalMotion and a 3-D Scene 363</p> <p>10.3 Image Alignment 365</p> <p>10.3.1 Feature-Based Methods 367</p> <p>10.3.2 Mechanical-Based Methods 369</p> <p>10.4 Image Composition 370</p> <p>10.4.1 Compositing Surface 370</p> <p>10.4.2 ImageWarping 371</p> <p>10.4.3 Pixel Selection and Blending 373</p> <p>10.5 Application Examples 374</p> <p>10.5.1 Pushbroom Stereo Mosaics Under TranslationalMotion 374</p> <p>10.5.1.1 Parallel-Perspective Geometry and Panoramas 374</p> <p>10.5.1.2 Stereo and Multiview Panoramas 376</p> <p>10.5.1.3 Results 378</p> <p>10.5.2 Stereo Mosaics when Moving a Camera on a Circular Path 378</p> <p>10.5.2.1 Circular Geometry 379</p> <p>10.5.2.2 Stereo Geometry 379</p> <p>10.5.2.3 Geometry and ResultsWhen Using PRISM 381</p> <p>10.5.3 Multimodal Panoramic Registration of Video Images 382</p> <p>10.5.3.1 Concentric Geometry 383</p> <p>10.5.3.2 Multimodal Alignment 385</p> <p>10.5.3.3 Results 387</p> <p>10.5.4 Video Mosaics Under GeneralMotion 387</p> <p>10.5.4.1 Direct Layering Approach 389</p> <p>10.5.4.2 Multiple Runs and Results 392</p> <p>10.6 Further Reading 393</p> <p>References 395</p> <p><b>11 Multitemporal Image Registration 397</b></p> <p>11.1 Introduction 397</p> <p>11.2 Finding Transformation Parameters from Line Parameters 398</p> <p>11.3 Finding an Initial Set of Homologous Lines 399</p> <p>11.4 Maximizing the Number of Homologous Lines 403</p> <p>11.5 Examples of Multitemporal Image Registration 406</p> <p>11.6 Further Reading 413</p> <p>References 415</p> <p><b>12 Open Problems and Research Topics 419</b></p> <p>12.1 Finding Rotational Difference between Multimodality Images 419</p> <p>12.2 Designing a Robust Image Descriptor 420</p> <p>12.3 Finding Homologous Lines for Nonrigid Registration 421</p> <p>12.4 Nonrigid Registration Using Homologous Lines 423</p> <p>12.5 Transformation Models with Nonsymmetric Basis Functions 423</p> <p>12.6 Finding Homologous Points along Homologous Contours 426</p> <p>12.7 4-D Image Registration 429</p> <p>References 430</p> <p>Glossary 433</p> <p>Acronyms 437</p> <p>Symbols 439</p> <p>A Image Registration Software 441</p> <p>A.1 Chapter 2: Image Orientation Detection 441</p> <p>A.1.1 Introduction 441</p> <p>A.1.2 Operations 442</p> <p>A.2 Chapter 3: Feature Point Detection 444</p> <p>A.2.1 Introduction 444</p> <p>A.2.2 Operations 445</p> <p>A.3 Chapter 4: Feature Line Detection 448</p> <p>A.3.1 Introduction 448</p> <p>A.3.2 Operations 449</p> <p>A.4 Chapter 5: Finding Homologous Points 452</p> <p>A.4.1 Introduction 452</p> <p>A.4.2 Operations 452</p> <p>A.5 Chapter 6: Finding Homologous Lines 459</p> <p>A.5.1 Introduction 459</p> <p>A.5.2 Operations 460</p> <p>A.6 Chapter 7: Nonrigid Image Registration 469</p> <p>A.6.1 Introduction 469</p> <p>A.6.2 Operations 469</p> <p>A.7 Chapter 8: Volume Image Registration 479</p> <p>A.7.1 Introduction 479</p> <p>A.7.2 I/O File Formats 479</p> <p>A.7.3 Operations 480</p> <p>References 487</p> <p>Index 489</p>

<p><strong>Arthur Ardeshir Goshtasby, PhD,</strong> is a professor in the Department of Computer Science and Engineering at Wright State University. Dr. Goshtasby has more than thirty years of experience in the areas of computer vision and pattern recognition and has published more than sixty journal articles and seven book chapters, addressing issues in image registration. He is the author of 2-D and 3-D Image Registration (Wiley, 2005).

<p><strong> A hands-on guide to image registration theory and methods—with examples of a wide range of real-world applications</strong> <p> <em>Theory and Applications of Image Registration</em> offers comprehensive coverage of feature-based image registration methods. It provides in-depth exploration of an array of fundamental issues, including image orientation detection, similarity measures, feature extraction methods, and elastic transformation functions. Also covered are robust parameter estimation, validation methods, multi-temporal and multi-modality image registration, methods for determining the orientation of an image, methods for identifying locally unique neighborhoods in an image, methods for detecting lines in an image, methods for finding corresponding points and corresponding lines in images, registration of video images to create panoramas, and much more. <p><em>Theory and Applications of Image Registration</em> provides readers with a practical guide to the theory and underpinning principles. Throughout the book numerous real-world examples are given, illustrating how image registration can be applied to problems in various fields, including biomedicine, remote sensing, and computer vision. Also provided are software routines to help readers develop their image registration skills. Many of the algorithms described in the book have been implemented, and the software packages are made available to the readers of the book on a companion website. In addition, the book: <ul> <li>Explores the fundamentals of image registration and provides a comprehensive look at its multi-disciplinary applications</li> <li>Reviews real-world applications of image registration in the fields of biomedical imaging, remote sensing, computer vision, and more</li> <li>Discusses methods in the registration of long videos in target tracking and 3-D reconstruction</li> <li>Addresses key research topics and explores potential solutions to a number of open problems in image registration</li> <li>Includes a companion website featuring fully implemented algorithms and image registration software for hands-on learning</li> </ul> <br> <p> <em>Theory and Applications of Image Registration</em> is a valuable resource for researchers and professionals working in industry and government agencies where image registration techniques are routinely employed. It is also an excellent supplementary text for graduate students in computer science, electrical engineering, software engineering, and medical physics.

US