Muutke küpsiste eelistusi

Processing, Analyzing and Learning of Images, Shapes, and Forms: Part 1, Volume 19 [Kõva köide]

Volume editor (Hong Kong Baptist University, Hong Kong, University of Berge, Norway), Volume editor (Technion - Israel Institute of Technology, Israel)
  • Formaat: Hardback, 157 pages, kõrgus x laius: 229x152 mm, kaal: 380 g
  • Sari: Handbook of Numerical Analysis
  • Ilmumisaeg: 08-Nov-2018
  • Kirjastus: North-Holland
  • ISBN-10: 0444642056
  • ISBN-13: 9780444642059
Teised raamatud teemal:
Processing, Analyzing and Learning of Images, Shapes, and Forms: Part 1, Volume 19Suurem pilt
  • Formaat: Hardback, 157 pages, kõrgus x laius: 229x152 mm, kaal: 380 g
  • Sari: Handbook of Numerical Analysis
  • Ilmumisaeg: 08-Nov-2018
  • Kirjastus: North-Holland
  • ISBN-10: 0444642056
  • ISBN-13: 9780444642059
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9780444642059.html
Märksõnad:

Processing, Analyzing and Learning of Images, Shapes, and Forms: Volume 19, Part One provides a comprehensive survey of the contemporary developments related to the analysis and learning of images, shapes and forms. It covers mathematical models as well as fast computational techniques, and includes new chapters on Alternating diffusion: a geometric approach for sensor fusion, Shape Correspondence and Functional Maps, Geometric models for perception-based image processing, Decomposition schemes for nonconvex composite minimization: theory and applications, Low rank matrix recovery: algorithms and theory, Geometry and learning for deformation shape correspondence, and Factoring scene layout from monocular images in presence of occlusion.

  • Presents a contemporary view on the topic, comprehensively covering the newest developments and content
  • Provides a comprehensive survey of the contemporary developments related to the analysis and learning of images, shapes and forms

Arvustused

"It ranges from a novel attempt to put deep learning within the framework of compressed sensing and sparse models, reconstruction of low rank matrices, shifting into learning geometry, shape representation that has the potential to migrate geometry analysis into that of deep learning, and pure geometric problems dealt in a novel, yet axiomatic, manner." --zbMATH

Contributors ix
Preface xi
Section One
1 Compressed Learning for Image Classification: A Deep Neural Network Approach
3(18)
E. Zisselman
A. Adler
M. Elad
1 Introduction
4(2)
2 Compressed Learning Overview
6(3)
2.1 Compressed Sensing
6(2)
2.2 Compressed Learning
8(1)
3 The Proposed End-to-End CL Approach
9(1)
4 Performance Evaluation
10(6)
4.1 MNIST Dataset
10(4)
4.2 CIFAR10 Dataset
14(2)
5 Conclusions
16(5)
References
16(5)
Section Two
2 Exploiting the Structure Effectively and Efficiently in Low-Rank Matrix Recovery
21(34)
Jian-Feng Cai
Ke Wei
1 Introduction
22(2)
1.1 Matrix Sensing
23(1)
1.2 Matrix Completion
23(1)
1.3 Phase Retrieval
23(1)
1.4 Notation and Organization
24(1)
2 Convex Approach: Nuclear Norm Minimization
24(7)
2.1 Recovery Guarantees of Nuclear Norm Minimization
25(2)
2.2 Algorithms for Nuclear Norm Minimization
27(4)
3 Projected Gradient Descent Based on Matrix Factorization
31(5)
3.1 Recovery Guarantees of (PGD)
32(4)
4 Algorithms on Embedded Manifold of Low-Rank Matrices
36(10)
4.1 Iterative Hard Thresholding
36(1)
4.2 Riemannian Optimization on Low-Rank Manifold
37(2)
4.3 Recovery Guarantees of RGrad
39(1)
4.4 PGD vs RGrad: An Illustration on Matrix Completion
40(1)
4.5 Extensions
41(5)
5 Conclusion and Discussion
46(9)
References
46(9)
Section Three
3 Partial Single- and Multishape Dense Correspondence Using Functional Maps
55(86)
Or Litany
Emanuele Rodola
Alex Bronstein
Michael Bronstein
Daniel Cremers
1 Introduction
56(2)
1.1 Related Work
56(2)
2 Full Nonrigid Shape Correspondence
58(4)
2.1 Functional Representation
59(1)
2.2 Joint Diagonalization
60(2)
3 Partial Functional Correspondence
62(2)
4 Multipart Partial Functional Maps
64(10)
4.1 Nonrigid Puzzles
65(3)
4.2 Implementation
68(1)
4.3 Experimental Results
69(3)
4.4 Discussion and Conclusions
72(2)
5 Fully Spectral Partial Functional Maps
74(17)
5.1 Localization
74(1)
5.2 Problem
75(2)
5.3 Part-to-Part
77(1)
5.4 Comparison to Joint Diagonalization
77(1)
5.5 Geometric Interpretation
78(1)
5.6 Implementation
79(2)
5.7 Experimental Results
81(1)
5.8 Part-to-Full
82(4)
5.9 Discussion and Conclusions
86(2)
References
88(3)
4 Shape Correspondence and Functional Maps
91(1)
Maks Ovsjanikov
List of Key Symbols
92(1)
1 Introduction
92(1)
1.1 Organization
93(1)
2 Functional Maps in the Continuous Setting
93(2)
3 Functional Maps in a Basis
95(3)
3.1 Functional Representation of Given Map
96(1)
3.2 General Functional Maps
97(1)
4 Functional Representation Properties
98(6)
4.1 Choice of Basis
98(2)
4.2 Map Properties in the Functional Domain
100(1)
4.3 Map Inversion and Composition
101(1)
4.4 Linearity of Constraints
101(1)
4.5 Operator Commutativity
102(1)
4.6 Descriptor Preservation via Commutativity
103(1)
5 Shape Matching With Functional Maps
104(9)
5.1 Basis Estimation
105(1)
5.2 Functional Map Optimization Problem
106(1)
5.3 Functional Map Computation and Regularization
106(2)
5.4 Efficient Conversion to Point-to-Point
108(1)
5.5 Postprocessing Iterative Refinement
108(1)
5.6 Shape Matching Pipeline
109(1)
5.7 Basic Implementation
110(1)
5.8 Results
111(2)
6 Other Applications and Extensions
113(2)
6.1 Function (Segmentation) Transfer
113(1)
6.2 Extensions
113(1)
6.3 Maps in Collections
114(1)
6.4 Functional Maps and Learning
114(1)
6.5 Coupled Functional Maps and Adjoint Regularization
114(1)
7 Further Reading
115(4)
Acknowledgements
116(1)
References
116(3)
5 Factoring Scene Layout From Monocular Images in Presence of Occlusion
119(1)
Niloy J. Mitra
1 Introduction
120(2)
2 Related Work
122(1)
2.1 Scene Mock-ups
122(1)
2.2 3D →2D Alignment
123(1)
2.3 Priors for Scene Reconstruction
123(1)
3 Overview
123(1)
4 Method
124(5)
4.1 Keypoint Detection
124(1)
4.2 Candidate Object Detection
124(2)
4.3 Scene Inference
126(3)
5 Results and Discussion
129(7)
5.1 Training and Test Data
129(1)
5.2 Performance Measures and Parameters
130(1)
5.3 Baselines: State-of-the-Art Alternatives
131(1)
5.4 Evaluation and Discussion
132(4)
5.5 Ablation Study
136(1)
6 Discussion
136(1)
7 Conclusion
137(4)
Acknowledgements
137(1)
References
137(4)
Index 141
Ron Kimmel is a Professor of Computer Science at the Technion where he holds the Montreal Chair in Sciences. He held a post-doctoral position at UC Berkeley and a visiting professorship at Stanford University. He has worked in various areas of image and shape analysis in computer vision, image processing, and computer graphics. Kimmel's interest in recent years has been non-rigid shape processing and analysis, medical imaging and computational biometry, numerical optimization of problems with a geometric flavor, and applications of metric geometry, deep learning, and differential geometry. Kimmel is an IEEE Fellow for his contributions to image processing and non-rigid shape analysis. He is an author of two books, an editor of one, and an author of numerous articles. He is the founder of the Geometric Image Processing Lab. and a founder and advisor of several successful image processing and analysis companies. Professor Tai Xue-Cheng is a member of the Department of Mathematics at the Hong Kong Baptist University, Hong Kong and also the University of Bergen of Norway. His research interests include Numerical partial differential equations, optimization techniques, inverse problems, and image processing. He is the winner for several prizes for his contributions to scientific computing and innovative researches for image processing. He served as organizing and program committee members for many international conferences and has been often invited for international conferences. He has served as referee and reviewers for many premier conferences and journals.