Regularization and Bayesian Methods for Inverse

Problems in Signal and Image Processing 1st

Edition Jean-Franã§Ois Giovannelli pdf download

https://ebookfinal.com/download/regularization-and-bayesian-
methods-for-inverse-problems-in-signal-and-image-processing-1st-
edition-jean-frana%c2%a7ois-giovannelli/

Explore and download more ebooks or textbooks

at ebookfinal.com
 We believe these products will be a great fit for you. Click
the link to download now, or visit ebookfinal
to discover even more!

MacMillan Destination C1 C2 Grammar Vocabulary Malcolm

Mann

https://ebookfinal.com/download/macmillan-destination-c1-c2-grammar-
vocabulary-malcolm-mann/

Nonlinear Signal and Image Processing Theory Methods and

Applications 1st Edition Kenneth E. Barner

https://ebookfinal.com/download/nonlinear-signal-and-image-processing-
theory-methods-and-applications-1st-edition-kenneth-e-barner/

Bayesian Signal Processing Classical Modern and Particle

Filtering Methods 2ed. Edition James V. Candy

https://ebookfinal.com/download/bayesian-signal-processing-classical-
modern-and-particle-filtering-methods-2ed-edition-james-v-candy/

Sparse Image and Signal Processing Wavelets and Related

Geometric Multiscale Analysis 2ed. Edition Jean-Luc Starck

https://ebookfinal.com/download/sparse-image-and-signal-processing-
wavelets-and-related-geometric-multiscale-analysis-2ed-edition-jean-
luc-starck/
Signal and image processing for remote sensing 2ed.
Edition Chen C.H. (Ed.)

https://ebookfinal.com/download/signal-and-image-processing-for-
remote-sensing-2ed-edition-chen-c-h-ed/

Computational methods for inverse problems 1st Edition

Curtis R. Vogel

https://ebookfinal.com/download/computational-methods-for-inverse-
problems-1st-edition-curtis-r-vogel/

Adaptive Blind Signal and Image Processing 1st Edition

Andrzej Cichocki

https://ebookfinal.com/download/adaptive-blind-signal-and-image-
processing-1st-edition-andrzej-cichocki/

Numerical methods for solving inverse problems of

mathematical physics 1st Edition Samarskii

https://ebookfinal.com/download/numerical-methods-for-solving-inverse-
problems-of-mathematical-physics-1st-edition-samarskii/

Digital Signal and Image Processing Using MATLAB Gérard

Blanchet

https://ebookfinal.com/download/digital-signal-and-image-processing-
using-matlab-gerard-blanchet/
Regularization and Bayesian Methods for Inverse
Problems in Signal and Image Processing 1st Edition
Jean-Franã§Ois Giovannelli Digital Instant Download
Author(s): Jean-François Giovannelli, JérÃ'me Idier
ISBN(s): 9781118827253, 1848216378
Edition: 1
File Details: PDF, 12.40 MB
Year: 2015
Language: english
DIGITAL SIGNAL AND IMAGE PROCESSING SERIES

Regularization and
Bayesian Methods
for Inverse Problems in
Signal and Image Processing

Edited by
Jean-François Giovannelli
Jérôme Idier
Regularization and Bayesian Methods for Inverse Problems
in Signal and Image Processing
 Series Editor
Henri Maître

Regularization and Bayesian

Methods for Inverse
Problems in Signal and
Image Processing

Edited by

Jean-François Giovannelli
Jérôme Idier
First published 2015 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the
CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:

ISTE Ltd John Wiley & Sons, Inc.

27-37 St George’s Road 111 River Street
London SW19 4EU Hoboken, NJ 07030
UK USA

www.iste.co.uk www.wiley.com

© ISTE Ltd 2015

The rights of Jean-François Giovannelli and Jérôme Idier to be identified as the authors of this work have
been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Control Number: 2014956810

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library
ISBN 978-1-84821-637-2
 Contents

I NTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Jean-François G IOVANNELLI and Jérôme I DIER

C HAPTER 1. 3D R ECONSTRUCTION IN X- RAY T OMOGRAPHY:

A PPROACH E XAMPLE FOR C LINICAL DATA P ROCESSING . . . . . . . . . 1
Yves G OUSSARD
1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1. Data formation models . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2. Estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.3. Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1. Data formation models . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2. Estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.3. Minimization method . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.4. Implementation of the reconstruction procedure . . . . . . . . . . . 14
1.4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.1. Comparison of minimization algorithms . . . . . . . . . . . . . . . . 15
1.4.2. Using a region of interest in reconstruction . . . . . . . . . . . . . . 18
1.4.3. Consideration of the polyenergetic character of the source . . . . . 21
1.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.6. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

CHAPTER 2. A NALYSIS OF F ORCE -VOLUME I MAGES IN ATOMIC

F ORCE M ICROSCOPY U SING S PARSE A PPROXIMATION . . . . . . . . . . 31
Charles S OUSSEN, David B RIE, Grégory F RANCIUS, Jérôme I DIER
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
vi Regularization and Bayesian Methods for Inverse Problems

2.2. Atomic force microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.2.1. Biological cell characterization . . . . . . . . . . . . . . . . . . . . . 32
2.2.2. AFM modalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.3. Physical piecewise models . . . . . . . . . . . . . . . . . . . . . . . 37
2.3. Data processing in AFM spectroscopy . . . . . . . . . . . . . . . . . . . 40
2.3.1. Objectives and methodology in signal processing . . . . . . . . . . 40
2.3.2. Segmentation of a force curve by sparse approximation . . . . . . . 41
2.4. Sparse approximation algorithms . . . . . . . . . . . . . . . . . . . . . . 43
2.4.1. Minimization of a mixed 2 -0 criterion . . . . . . . . . . . . . . . . 44
2.4.2. Dedicated algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.3. Joint detection of discontinuities . . . . . . . . . . . . . . . . . . . . 46
2.5. Real data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5.1. Segmentation of a retraction curve:
comparison of strategies . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5.2. Retraction curve processing . . . . . . . . . . . . . . . . . . . . . . . 50
2.5.3. Force-volume image processing in the approach phase . . . . . . . 52
2.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

C HAPTER 3. P OLARIMETRIC I MAGE R ESTORATION BY N ON - LOCAL

M EANS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Sylvain FAISAN, François ROUSSEAU, Christian H EINRICH, Jihad Z ALLAT
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2. Light polarization and the Stokes–Mueller formalism . . . . . . . . . . 58
3.3. Estimation of the Stokes vectors . . . . . . . . . . . . . . . . . . . . . . 61
3.3.1. Estimation of the Stokes vector in a pixel . . . . . . . . . . . . . . . 61
3.3.2. Non-local means filtering . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3.3. Adaptive non-local means filtering . . . . . . . . . . . . . . . . . . . 66
3.3.4. Application to the estimation of Stokes vectors . . . . . . . . . . . . 69
3.4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.1. Results with synthetic data . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.2. Results with real data . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.6. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

C HAPTER 4. V IDEO P ROCESSING AND R EGULARIZED I NVERSION

M ETHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Guy L E B ESNERAIS, Frédéric C HAMPAGNAT
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2. Three applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2.1. PIV and estimation of optical flow . . . . . . . . . . . . . . . . . . . 82
4.2.2. Multiview stereovision . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.3. Superresolution and non-translational motion . . . . . . . . . . . . . 86
 Contents vii

4.3. Dense image registration . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.3.1. Direct formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.3.2. Variational formulation . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3.3. Extension of direct formulation for multiview processing . . . . . . 92
4.4. A few achievements based on direct formulation . . . . . . . . . . . . . 92
4.4.1. Dense optical flow by correlation of local window . . . . . . . . . . 92
4.4.2. Occlusion management in multiview stereovision . . . . . . . . . . 97
4.4.3. Direct models for SR . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.6. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

C HAPTER 5. BAYESIAN A PPROACH IN P ERFORMANCE M ODELING :

A PPLICATION TO S UPERRESOLUTION . . . . . . . . . . . . . . . . . . . . . 109
Frédéric C HAMPAGNAT, Guy L E B ESNERAIS, Caroline K ULCSÁR
5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.1.1. The hiatus between performance modeling and Bayesian inversion 109
5.1.2. Chapter organization . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.2. Performance modeling and Bayesian paradigm . . . . . . . . . . . . . . 111
5.2.1. An empirical performance evaluation tool . . . . . . . . . . . . . . . 111
5.2.2. Usefulness and limits of a performance evaluation tool . . . . . . . 111
5.2.3. Bayesian formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3. Superresolution techniques behavior . . . . . . . . . . . . . . . . . . . . 113
5.3.1. Superresolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.3.2. SR methods performance: known facts . . . . . . . . . . . . . . . . 115
5.3.3. An SR experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3.4. Performance model and properties . . . . . . . . . . . . . . . . . . . 122
5.4. Application examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.4.1. Behavior of the optimal filter with regard to the number of images . 127
5.4.2. Characterization of an approximation: shifts rounding . . . . . . . . 129
5.5. Real data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.5.1. A concrete measure to improve the resolution: the RER . . . . . . . 132
5.5.2. Empirical validation and application field . . . . . . . . . . . . . . . 134
5.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

C HAPTER 6. L INE S PECTRA E STIMATION FOR I RREGULARLY

S AMPLED S IGNALS IN A STROPHYSICS . . . . . . . . . . . . . . . . . . . . . 141
Sébastien B OURGUIGNON, Hervé C ARFANTAN
6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2. Periodogram, irregular sampling, maximum likelihood . . . . . . . . . 144
6.3. Line spectra models: spectral sparsity . . . . . . . . . . . . . . . . . . . 146
6.3.1. An inverse problem with sparsity prior information . . . . . . . . . 147
6.3.2. Difficulties in terms of sparse approximation . . . . . . . . . . . . . 149
viii Regularization and Bayesian Methods for Inverse Problems

6.4. Prewhitening, CLEAN and greedy approaches . . . . . . . . . . . . . . 151

6.4.1. Standard greedy algorithms . . . . . . . . . . . . . . . . . . . . . . . 151
6.4.2. A more complete iterative method: single best replacement . . . . . 153
6.4.3. CLEAN-based methods . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.5. Global approach and convex penalization . . . . . . . . . . . . . . . . . 155
6.5.1. Significance of 1 penalization in . . . . . . . . . . . . . . . . . . 156
6.5.2. Existence and uniqueness . . . . . . . . . . . . . . . . . . . . . . . . 156
6.5.3. Minimizer and regularization parameter characterization . . . . . . 157
6.5.4. Amplitude bias and a posteriori corrections . . . . . . . . . . . . . . 157
6.5.5. Hermitian symmetry and specificity of the zero frequency . . . . . 158
6.5.6. Optimization algorithms . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.5.7. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.6. Probabilistic approach for sparsity . . . . . . . . . . . . . . . . . . . . . 159
6.6.1. Bernoulli–Gaussian model for spectral analysis . . . . . . . . . . . 160
6.6.2. A structure adapted to the use of MCMC methods . . . . . . . . . . 161
6.6.3. An extended BG model for improved accuracy . . . . . . . . . . . . 162
6.6.4. Stochastic simulation and estimation . . . . . . . . . . . . . . . . . . 162
6.6.5. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
6.8. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

C HAPTER 7. J OINT D ETECTION -E STIMATION IN

F UNCTIONAL MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Philippe C IUCIU, Florence F ORBES, Thomas V INCENT, Lotfi C HAARI

7.1. Introduction to functional neuroimaging . . . . . . . . . . . . . . . . . . 169

7.2. Joint detection-estimation of brain activity . . . . . . . . . . . . . . . . 171
7.2.1. Detection and estimation: two interdependent issues . . . . . . . . . 171
7.2.2. Hemodynamics physiological hypotheses . . . . . . . . . . . . . . . 173
7.2.3. Spatially variable convolutive model . . . . . . . . . . . . . . . . . . 175
7.2.4. Regional generative model . . . . . . . . . . . . . . . . . . . . . . . 176
7.3. Bayesian approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
7.3.1. Likelihood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
7.3.2. A priori distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
7.3.3. A posteriori distribution . . . . . . . . . . . . . . . . . . . . . . . . . 182
7.4. Scheme for stochastic MCMC inference . . . . . . . . . . . . . . . . . . 183
7.4.1. HRF and NRLs simulation . . . . . . . . . . . . . . . . . . . . . . . 183
7.4.2. Unsupervised spatial and spatially adaptive regularization . . . . . 184
7.5. Alternative variational inference scheme . . . . . . . . . . . . . . . . . . 184
7.5.1. Motivations and foundations . . . . . . . . . . . . . . . . . . . . . . 184
7.5.2. Variational EM algorithm . . . . . . . . . . . . . . . . . . . . . . . . 186
7.6. Comparison of both types of solutions . . . . . . . . . . . . . . . . . . . 190
7.6.1. Experiments on simulated data . . . . . . . . . . . . . . . . . . . . . 190
7.6.2. Experiments on real data . . . . . . . . . . . . . . . . . . . . . . . . 193
 Contents ix

7.7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

7.8. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

C HAPTER 8. MCMC AND VARIATIONAL A PPROACHES FOR BAYESIAN

I NVERSION IN D IFFRACTION I MAGING . . . . . . . . . . . . . . . . . . . . 201
Hacheme AYASSO, Bernard D UCHÊNE, Ali M OHAMMAD -D JAFARI
8.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
8.2. Measurement configuration . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.2.1. The microwave device . . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.2.2. The optical device . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
8.3. The forward model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
8.3.1. The microwave case . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
8.3.2. The optical case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
8.3.3. The discrete model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.3.4. Validation of the forward model . . . . . . . . . . . . . . . . . . . . 210
8.4. Bayesian inversion approach . . . . . . . . . . . . . . . . . . . . . . . . 211
8.4.1. The MCMC sampling method . . . . . . . . . . . . . . . . . . . . . 213
8.4.2. The VBA method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
8.4.3. Initialization, progress and convergence of the algorithms . . . . . . 217
8.5. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

C HAPTER 9. VARIATIONAL BAYESIAN A PPROACH AND B I -M ODEL

FOR THE R ECONSTRUCTION -S EPARATION OF A STROPHYSICS
C OMPONENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Thomas RODET, Aurélia F RAYSSE, Hacheme AYASSO
9.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
9.2. Variational Bayesian methodology . . . . . . . . . . . . . . . . . . . . . 228
9.3. Exponentiated gradient for variational Bayesian . . . . . . . . . . . . . 229
9.4. Application: reconstruction-separation of astrophysical
components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
9.4.1. Direct model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
9.4.2. A priori distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
9.4.3. A posteriori distribution . . . . . . . . . . . . . . . . . . . . . . . . . 235
9.5. Implementation of the variational Bayesian approach . . . . . . . . . . 236
9.5.1. Separability study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
9.5.2. Update of the approximation distributions . . . . . . . . . . . . . . . 236
9.6. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
9.6.1. Simulated data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
9.6.2. Real data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
9.7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
9.8. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
x Regularization and Bayesian Methods for Inverse Problems

C HAPTER 10. K ERNEL VARIATIONAL A PPROACH FOR TARGET

T RACKING IN A W IRELESS S ENSOR N ETWORK . . . . . . . . . . . . . . . 251
Hichem S NOUSSI, Paul H ONEINE, Cédric R ICHARD
10.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
10.2. State of the art: limitations of existing methods . . . . . . . . . . . . . 252
10.3. Model-less target tracking . . . . . . . . . . . . . . . . . . . . . . . . . 254
10.3.1. Construction of the likelihood by matrix regression . . . . . . . . . 255
10.3.2. Variational filtering for the tracking of mobile objects . . . . . . . 258
10.4. Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
10.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
10.6. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264

C HAPTER 11. E NTROPIES AND E NTROPIC C RITERIA . . . . . . . . . . . 267

Jean-François B ERCHER
11.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
11.2. Some entropies in information theory . . . . . . . . . . . . . . . . . . . 268
11.2.1. Main properties and definitions . . . . . . . . . . . . . . . . . . . . 268
11.2.2. Entropies and divergences in the continuous case . . . . . . . . . . 270
11.2.3. Maximum entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
11.2.4. Escort distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
11.3. Source coding with escort distributions
and Rényi bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
11.3.1. Source coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
11.3.2. Source coding with Campbell measure . . . . . . . . . . . . . . . . 274
11.3.3. Source coding with escort mean . . . . . . . . . . . . . . . . . . . 275
11.4. A simple transition model . . . . . . . . . . . . . . . . . . . . . . . . . 277
11.4.1. The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
11.4.2. The Rényi divergence as a consequence . . . . . . . . . . . . . . . 279
11.4.3. Fisher information for the parameter q . . . . . . . . . . . . . . . . 279
11.4.4. Distribution inference with
generalized moment constraint . . . . . . . . . . . . . . . . . . . . . 281
11.5. Minimization of the Rényi divergence and associated entropies . . . 281
11.5.1. Minimization under generalized moment constraint . . . . . . . . 282
11.5.2. A few properties of the partition functions . . . . . . . . . . . . . . 283
11.5.3. Entropic functionals derived from the Rényi divergence . . . . . . 285
11.5.4. Entropic criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
11.6. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

L IST OF AUTHORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293

I NDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
 Introduction

This book was written in tribute to our colleague Guy Demoment, who was a
researcher at the CNRS from 1977 to 1988, then Professor at the University of Paris-
Sud until 2008, member of the Laboratoire des Signaux et Systèmes (L2S, UMR 8506,
Gif-sur-Yvette) and its director from 1997 to 2001, and the founder of a research group
on inverse problems in signal and image processing at the beginning of the 1980s.

Guy Demoment’s research activities began in 1970, at the interface between

biological and medical engineering, automatic control and the still fledgling field of
signal processing. Guy was particularly interested in cardiac function and in the
cardiovascular system [DEM 77]. He derived a mathematical model of the
functioning of the cardiovascular hemodynamic loop which was subsequently used to
develop the control law of cardiac replacement prostheses. He also focused on
aspects closer to theoretical biology such as left ventricle modeling and questions
closer to physics such as the determination of vascular impedance [DEM 81].

This latter aspect naturally leads us to confront models with reality by means of
measurements. The idea makes sense whereas in practice these measurements
provide only indirect and degraded information on the quantities of interest. These
degradations are generally considered in two forms: structure (resolution limitations,
dynamics, sampling, etc.) and uncertainty (measurement noise, model
approximation, etc.). The restitution of the quantity of interest then raises a real
ill-posed inversion or inference problem. By creating the Groupe Problèmes Inverses
(GPI – Inverse Problems Group), Guy promoted this scientific approach within the
L2S then to the whole of the signal-image community within the engineering
sciences. Having shared this approach with him within the GPI is a fortunate

Introduction written by Jean-François G IOVANNELLI and Jérôme I DIER.

xii Regularization and Bayesian Methods for Inverse Problems

opportunity that most of the co-authors of this book have benefited from, as doctoral
students or beginner colleagues.

Undoubtedly, Guy Demoment was an essential contributor to the field of inverse

problems in signal and image processing, and its main instigator in the French
community. He has also been passionate about related issues such as that of the
effective implementation of a number of algorithms. In particular, in the context of
linear deconvolution and adaptive spectral analysis, Kalman filtering and smoothing
algorithms were given particular attention. Guy made several significant
contributions concerning fast versions of these algorithms [DEM 85, DEM 91].

The exploitation of probabilistic models for detection-estimation has also been a

subject of choice for Guy and his collaborators since the end of the 1980s, resulting
in recursive [GOU 89], and then iterative [GOU 90] computational structures. It is
interesting to note that the latter are very competitive precursors to the well-known
greedy algorithms in parsimonious approximation, as is clearly stated in the book.

Regarding more fundamental subjects, he has been interested in the issues of

information modeling and Bayesian inference, inspired by E.T. Jaynes’ works. He
contributed to the use of a maximum entropy principle for the synthesis of a priori
models and to their application to tomography [MOH 87, MOH 88a], and then he
studied the principle of maximum entropy on the mean in the context of inverse
problems [LEB 99]. He then further explored these issues and, during his last period
of scientific activity, became interested in variational approaches for Bayesian
inference.

Guy’s scientific sensitivity has also been visible in a significant way in his teaching
activities. He created several courses ranging from undergraduate to PhD levels, as
well as in continuing education, and always dedicated to them a lot of energy and
creativity. Some particular examples include a course on Bayesian inference and the
basics of probabilities, and among the most in-depth themes, Kalman algorithms and
their fast versions, as well as the deconvolution of signals.

Beyond his scientific activities, researches and teachings, Guy has also been
involved in a remarkable way in community life. On a national level, he has been a
member of the Conseil National des Universités (www.cpcnu.fr), a particularly active
member of scholar and research networks, e.g. club EEA (www.clubeea.org) and the
GdR ISIS (gdr-isis.fr). Within the University of Paris-Sud, he has chaired the
pedagogy commission, he has been vice president of the Department of Physics in
Orsay, responsible for bachelor-level diploma, and co-creator of a masters-level
diploma.

With regard to the present book, it concerns “ill-posed inverse problems”. The
readers can refer to the widely cited article [DEM 89] or to a previous collective
 Introduction xiii

book [IDI 08] on this subject, of which Guy Demoment is one of the main
contributors. It is concerned with problems that cannot be resolved on the basis of the
observed data only and the construction of solutions requires other information,
referred to as a priori. These solutions are then specific to the information taken into
account. The recognition and the explanation of this information are necessary to
appreciate the range of validity and the scope of application of the constructed
solutions. Over the 1980s, the scientific community has greatly acknowledged the
significance of this problematic, and contributions have become very abundant not
only in the signal-image community but also in that of mathematics, computer
science and physics.

As a direct response to this thematic abundance concerning inverse problems, we

have chosen to address a broad spectrum of data processing problems and application
domains, with a particular focus on the diversity of mathematical tools.

From the point of view of application fields, an important part of the book is
dedicated to different scientific fields, which present a large number of inversion
problems: biological and medical imaging, and more specifically X-ray tomography
(Chapters 1, 2 and 7), astronomy (Chapters 6 and 9) as well as non-destructive
evaluation (Chapter 8). At least one other has been added: video sequence processing
(Chapters 4 and 5). Two other applications that are more rarely met in the field of
inversion: target tracking and sensor networks (Chapter 10) as well as digital
communications (Chapter 11).

The diversity of chapters is also evident when the considered acquisition

modalities come under scrutiny: from the more traditional ones such as tomography
(Chapters 1 and 8) and MRI (Chapter 7), optical imaging in the visible (Chapters 4
and 5) or in the infrared spectrum (Chapters 5 and 9) to more recent modalities such
as atomic force imaging (Chapter 2) and polarized optical imaging (Chapter 3).

Throughout the chapters, the duality between the approaches known as

“energetic” and “probabilistic” emerges. The first type of approach is based on
deterministic construction leading to criteria and to numerical optimization issues as
typically in Chapters 1, 2, 3, 4 and 6. The second type of approach is based on a
Bayesian construction, often hierarchical, which probabilizes unknown objects in
addition to data. It thus leads to a joint distribution; therefore, optimal strategies are
available and performance characterization right from the start becomes possible as
in Chapter 5. The remaining chapters make use of a posteriori distributions to
generate an estimation: they are explored by using stochastic sampling as in
Chapters 6, 7 and 8 or by an approximated maximization as in Chapters 7, 8, 9
and 10. The latter also introduces a notion of learning and relies on informational
principles discussed in Chapter 11, which presents more theoretical aspects related to
entropy criteria.
xiv Regularization and Bayesian Methods for Inverse Problems

I.1. Bibliography

[DEM 72] D EMOMENT G., Modèle de la boucle cardiovasculaire: évaluation de

l’autorégulation mécanique et de la fonction ventriculaire gauche, PhD thesis, no.169,
Orsay center, University of Paris-sud, 29 June 1972.
[DEM 77] D EMOMENT G., Contribution à l’étude du fonctionnement ventriculaire gauche
par des méthodes d’identification paramétriques. obtention d’un observateur de l’état du
ventricule, PhD thesis, no.1810, Orsay center, University of Paris-sud, 15 March 1977.
[DEM 82] D EMOMENT G., Introduction à la statistique, Lecture notes, École supérieure
d’électrité, no. 2906, 1982.
[DEM 83] D EMOMENT G., Déconvolution des signaux, Lecture notes, École supérieure
d’électrité no. 2964, 1983.
[DEM 85] D EMOMENT G., R EYNAUD R., “Fast minimum-variance deconvolution”, IEEE
Transactions on Acoustics, Speech and Signal Processing, vol. ASSP-33, pp. 1324–1326,
1985.
[DEM 87] D EMOMENT G., Algorithme rapides, Lecture notes, École supérieure d’électrité,
no. 3152, 1987.
[DEM 89a] D EMOMENT G., “Equations de Chandrasekhar et algorithmes rapides pour le
traitement du signal et des images”, Traitement du Signal, vol. 6, pp. 103–115, 1989.
[DEM 89b] D EMOMENT G., “Image reconstruction and restoration: Overview of common
estimation structures and problems”, IEEE Transactions on Acoustics, Speech and Signal
Processing, vol. ASSP-37, no. 12, pp. 2024–2036, December 1989.
[DEM 91] D EMOMENT G., R EYNAUD R., “Fast RLS algorithms and Chandrasekhar
equations”, H AYKIN S., (ed.), SPIE Conference on Adaptive Signal Processing, San Diego,
CA, pp. 357–367, July 1991.
[DEM 05a] D EMOMENT G., Probabilités: modélisation des incertitudes, inférence logique,
et traitement des données expérimentales. Deuxième partie: application au traitement du
signal, Lecture notes, University of Paris-sud, Orsay center, 2005.
[DEM 05b] D EMOMENT G., Probabilités: modélisation des incertitudes, inférence logique, et
traitement des données expérimentales. Première partie: bases de la théorie, Lecture notes,
University of Paris-sud, Orsay center, 2005.
[GOU 89] G OUSSARD Y., D EMOMENT G., “Recursive deconvolution of Bernoulli-Gaussian
processes using a MA representation”, IEEE Transactions on Geoscience and Remote
Sensing, vol. GE-27, pp. 384–394, 1989.
[GOU 90] G OUSSARD Y., D EMOMENT G., I DIER J., “A new algorithm for iterative
deconvolution of sparse spike trains”, IEEE International Conference on Acoustic, Speech
and Signal Processing, Albuquerque, NM, pp. 1547–1550, April 1990.
[IDI 08] I DIER J., (ed.), Bayesian Approach to Inverse Problems, ISTE, London and John
Wiley & Sons, New York, April 2008.
[LEB 99] L E B ESNERAIS G., B ERCHER J.-F., D EMOMENT G., “A new look at entropy for
solving linear inverse problems”, IEEE Transactions on Information Theory, vol. 45, no. 5,
pp. 1565–1578, July 1999.
 Introduction xv

[MOH 87] M OHAMMAD -D JAFARI A., D EMOMENT G., “Maximum entropy Fourier
synthesis with application to diffraction tomography”, Applied Optics, vol. 26, no. 10,
pp. 1745–1754, 1987.
[MOH 88a] M OHAMMAD -D JAFARI A., D EMOMENT G., “Maximum entropy reconstruction
in X ray and diffraction tomography”, IEEE Transactions on Medical Imaging, vol. MI-7,
no. 4, pp. 345–354, 1988.
[MOH 88b] M OHAMMAD -D JAFARI A., D EMOMENT G., “Utilisation de l’entropie dans les
problèmes de restauration et de reconstruction d’images”, Traitement du Signal, vol. 5,
no. 4, pp. 235–248, 1988.
 1

3D Reconstruction in X-ray
Tomography: Approach Example
for Clinical Data Processing

1.1. Introduction

Works presented in this chapter stem from three-dimensional (3D) reconstruction

problems, in X-ray computed tomography (XRCT), within a clinical framework.
More specifically, the practical objective was to use XRCT to detect and quantify
possible restenosis occurring in some patients after the insertion of a stent. The
quality of reconstructions achieved by clinical tomographs being insufficient for this
purpose, the concern was thus to develop a method capable of rebuilding small
structures in the presence of metal objects, in a 3D context, with a precision higher
than that of tomographs available in hospitals; in addition, this method was supposed
to work with computers commonly available in most research laboratories (such as
personal computers (PCs), without any particular architecture or processor
hardware).

The development of a solution clearly falls within the framework of conventional

inverse problem solving. However, it is essential to take into account the
characteristics of 3D XRCT, and notably the very large volume of data to be
processed, the geometric complexity of the collection process of raw data, and
practical barriers to access these data. In order to achieve the objective stated above,
difficulties are twofold: (1) at the methodological level, the development of an
inversion method adapted to the intrinsic characteristics of the problem to be
addressed and, (2) at the implementation level, accounting for the practical obstacles
mentioned above as well as for constraints on the processing time inherent in any

Chapter written by Yves G OUSSARD.

2 Regularization and Bayesian Methods for Inverse Problems

clinical application. In the following, we present the retained approach based on the
analysis of the main factors likely to improve the quality of reconstructions while
satisfying the practical constraints which we must face; this brings us to putting the
methodological aspects into perspective, with regard to practical questions, in light of
the applied objective of these works.

1.2. Problem statement

Although image reconstruction methods used in the first tomographs were of the
analytical type [AMB 73, HOU 73], the advantages of approaches based on
estimation [HER 71, HER 73, HER 76b], then the ill-posed nature of tomographic
reconstruction problems [HER 76a, HER 79] were recognized very early on. Over
the past 35 years, many academic studies focusing on tomographic reconstruction
have been carried out in the context of solving inverse problems. Generally, the
emphasis is on the main three elements of this type of approach, that is to say,
modeling of the data formation process, choice of the estimator, and development of
techniques that enable the practical computation of the estimate [DEM 89]. These
works have been partly customized according to various imaging modalities (for
example, transmission [HER 76a], emission [LEV 87], diffraction [BER 97]
tomography, and more recently optical and/or multiphysics tomography (see
[BOA 04] for a partial synthesis)) which present largely variable degrees of
difficulty: if estimation conditions are often very unfavorable in diffraction
tomography (eddy current tomography [TRI 10], seismic imaging [VAU 11]) due to
the strong non-linearity of underlying physical phenomena, to the importance of
attenuation phenomena and to the small number of observations with respect to the
number of unknowns, the inversion conditions are generally better in emission
tomography (SPECT for example) and can be qualified as relatively favorable in
XRCT. This explains why, in this area, reconstruction methods known as “naive”
provide results that have been used in clinical practice for several decades.
Thereafter, we present the elements likely to have a significant impact on the
performance of an XRCT inversion method.

1.2.1. Data formation models

In XRCT, all data formation models are based on the Beer–Lambert law, which
describes the attenuation of an X-ray beam through a medium whose spatial
distribution of the attenuation coefficients is referred to by μ. It takes the form:
   
N : P n0 exp − μ(s) ds [1.1]
D
Discovering Diverse Content Through
Random Scribd Documents
 By built-up
sections.

The above diagram gives in compact form the classification of

tunnels according to materials penetrated and methods of
excavation adopted, which have been described more fully in the
succeeding paragraphs. It may be noted here again that this is a
purely arbitrary classification, and serves mostly as a convenience in
discussing the different classes of tunnels without confusion.
 CHAPTER V.

METHODS OF TIMBERING OR STRUTTING TUNNELS.

The purpose of timbering or strutting in tunnel work is to prevent

the caving-in of the roof and side walls of the excavation previous to
the construction of the lining. As the strutting has to resist all the
pressures developed in the roof and side walls, which may be
exceedingly troublesome and of great intensity in loose soils, its
design and erection call for particular care. The method of strutting
adopted depends upon the method of excavation employed; but in
every case the problem is not only to build it strong enough to
withstand the pressures developed, but to do this as economically as
possible, and with as little hindrance as may be to the work which is
going on simultaneously and which will come later. Only the latter
general problems of strutting peculiar to all methods of tunnel work
will be considered here. For this consideration strutting may be
classified according to the material of which it is built, under the
heads of timber structures and iron structures.

Fig. 17.—Joining Tunnel Struts by

Halving.
 TIMBER STRUTTING.

Timber is nearly always employed for strutting

in tunnel work. So long as it has the requisite
strength, any kind of timber is suitable for
strutting, since, it being only temporarily
employed, its durability is a matter of slight
importance. Timber with good elastic properties,
like pine or spruce, is preferably chosen, since it
yields gradually under stress, thus warning the
engineer of the approach of danger; while oak
and other strong timbers resist until the last
moment, and then yield suddenly under the
Fig. 18.—Round
breaking load. Soft woods, moreover, are usually Timber Post and
lighter in weight than hard woods, which is a Cap Bearing.
considerable advantage where so much handling
is required in a restricted space. Round timbers are generally
employed, since they are less expensive, and quite as satisfactory in
other respects as sawed timbers. In the English and Austrian
methods of strutting, which are described further on, a few of the
principal struts are of sawed timbers.
The various timbers of the strutting are seldom attached by
framed joints, but wedges are used to give them the necessary
bearing against each other. Where framed joints are employed they
are made of the simplest form usually by halving the joining timbers,
as shown by Fig. 17. Fig. 18 shows a form of joint used where round
posts carry beams of similar shape. The reason why it is possible to
do away with jointed connections to such a great extent, is that the
strains which the timbers have to resist are either compressive or
bending strains, and because the timbers are so short that they do
not require to be spliced.
Strutting of Headings.—The method of strutting the heading
that is employed depends upon the material through which the
heading is driven. In solid rock strutting may not be required at all,
or only for the purpose of preventing the fall of loose blocks from
the roof, then vertical props are erected where required, or
horizontal beams are inserted into the side walls, as shown by Fig.
19. These horizontal beams may be used singly at dangerous places,
or they may be placed from 2 ft. to 3 ft. apart all along the heading.
In the latter case they usually carry a lagging of planks, which may
be placed at intervals or close together, and filled above with stone
in case the roof of the excavation is very unstable. Planks used in
this manner are usually called poling-boards. Where the side walls as
well as the roof require support, vertical side posts are employed to
carry the roof beams, as shown by Fig. 20; and, when necessary,
poling-boards are inserted between these posts and the walls of the
excavation.

Fig. 19.—Ceiling Strutting for Fig. 20.—Ceiling Strutting with

Tunnel Roofs. Side Post Supports.
 Fig. 21.—Sill, Side Post and Fig. 22.—Reinforced Cross
Cap Cross Frame Strutting. Frame Strutting for
Treacherous Materials.

Frame Strutting.—In very loose soils not only the roof and side
walls, but also the floor of the heading require strutting. In these
cases frame strutting is employed, as shown by Fig. 21. It consists
simply of a rectangular frame; at the top there is a crown bar
supported by two vertical side posts setting on a sill laid across the
bottom of the heading. These frames are spaced at close intervals,
and carry longitudinal planks or poling-boards. The sill of the frame
is sometimes omitted when the soil is stable enough to permit it,
and in its place wooden footing blocks are substituted to carry the
side posts. In soils where the pressures are great enough to bend
the crown bar, a secondary frame is employed, as shown by Fig. 22,
the two inclined roof members, or rafters, of which support the
crown bar at the center.
 Fig. 23.—Longitudinal Poling- Fig. 24.—Transverse Poling-
Board System of Roof Strutting. Board System of Roof Strutting.

It is the more common practice in driving headings through soft

soils to use inclined poling-boards to support the roof. Fig. 23 shows
one method of doing this. The method of operation is as follows:
Assuming the poling-boards a and b to be in place, and supported
by the frames A, B, C, as shown, the first step in continuation of the
work is to insert the poling-board c over the crown bar of frame C,
and under the block m. Excavation is then begun at the top, and as
fast as the soil is removed ahead of it the poling-board c is driven
ahead until its rear end only slightly overhangs the crown bar of
frame C. The remainder of the face of the heading is then excavated
nearly to the front end of the poling-board c, and another frame is
set up. By a succession of these operations the heading is advanced.
The poling-boards at the sides of the heading are placed in a similar
manner to the roof poling-boards. A second method of using inclined
poling-boards is shown by Fig. 24. Here the poling-boards run
transversely, and are supported by the arrangement of timbering
shown. The chief advantage of using these inclined poling-boards,
particularly in the manner shown by Fig. 23, is that the excavators
work under cover at all times, and are thus safe from falling
fragments or sudden cavings.
Box Strutting.—In very treacherous soils, such as quicksand, peat,
and laminated clay, box strutting is commonly employed. The
method of building this strutting is to set up at the face of the work
a rectangular frame, and use it as a guide in driving a lagging or
boxing of horizontal planks into the soft soil ahead. These planks
have sharp edges, and are driven to a distance of 2 ft. or 3 ft. into
the face of the heading, so as to inclose a rectangular body of earth.
This earth is excavated nearly to the ends of the planks, and then
another frame is inserted close up against the new face of the
excavation, which supports the planks so that the remainder of the
earth included by them may be removed. These two frames, with
their plank lagging, constitute a “box;” and a series of these boxes,
one succeeding another, form the strutting of the heading.
Strutting the Face.—In some cases it is found necessary to strut
the face of the heading in order to prevent it from caving in. This is
generally done by setting plank vertically, and bracing them up by
means of inclined props whose feet abut against the sill of the
nearest cross frame. This strutting is erected while the workmen are
placing the side and roof strutting, and is removed to permit
excavation.
Full Section Timber Strutting.—For strutting the full section
two forms of timbering are employed, known as the polygonal
system and the longitudinal system.
Longitudinal strutting consists of a timber structure so arranged
as to have all the principal members supporting the poling-boards
parallel to the axis of the tunnel. This system of strutting is peculiar
to the English method of tunneling. The longitudinal timbers rest on
this finished masonry at one end, and are carried on a cross frame
or by props at the other end. At intermediate points the longitudinals
are braced apart by struts in planes transverse to the tunnel axis.
This construction makes a very strong strutting framework, since the
transverse struts act as arch ribs to stiffen the longitudinals; but the
use of transverse poling-boards requires the excavation of a larger
cross-section than is necessary when longitudinal poling-boards are
employed, and this increases the cost both for the amount of earth
excavated and the greater quantity of filling required.
In polygonal strutting the main members are in a plane normal to
the axis of the tunnel. They form a polygon whose sides follow
closely the sectional profile of the excavation. These polygonal
frames are placed at more or less short intervals apart, and are
braced together by short longitudinal struts lying close to the sides
of the excavation, and running from one frame to the next, and also
by longer longitudinal members which extend over several frames.
The polygonal system of strutting is peculiar to the Austrian method
of tunneling, and is fully described in a succeeding chapter. One of
its distinctive characteristics is that the poling-boards are inserted
parallel to the tunnel axis. Polygonal strutting is generally held to be
stronger than longitudinal strutting under uniform loads, but it is
more liable to distortion when the loads are unsymmetrical.

Fig. 25.—Shaft with Single Transverse Strutting.

 Fig. 26.—Rectangular Frame Strutting for Shafts.

Fig. 27.—Reinforced Rectangular Frame Strutting for Shafts in Treacherous

Materials.

Strutting of Shafts.—Tunnel shafts are strutted both to prevent

the caving-in of the sides and to divide them into compartments.
When the material penetrated is very compact, and caving is not
likely, a single series of transverse struts, one above the other,
running from the top to the bottom of the shaft, as shown by Fig.
25, is used to divide it into two compartments. In softer material,
where the sides of the shaft require support, Fig. 26 shows a form of
strutting commonly employed. It consists of vertical corner posts
braced apart at intervals by four horizontal struts placed close to the
walls of the shaft. The longer side struts are also braced apart at the
center by a middle strut which divides the shaft into two
compartments. A lagging of vertical plank is placed between the
walls of the shaft and the horizontal side struts. In very loose soils
the form of strutting shown by Fig. 27 is employed. This is practically
the same construction as is shown by Fig. 26, with the addition of an
interior polygonal horizontal bracing in each half of the shaft.
Referring to Fig. 27, the timbers a, a, etc., are vertical and
continuous from the top to the bottom of the shaft; and the
horizontal timbers, b, b, etc., are spaced at more or less close
intervals vertically. The lagging planks may be laid with spaces
between them, or close together, or, in case of very loose material,
with their edges overlapping. The manner of constructing the
strutting is also governed by the stability of the soil. In firm soils it is
possible to sink the shaft quite a depth without timbering, and the
timbering can be erected in sections of considerable length, which is
always an advantage, but in loose soils the timbering has to follow
closely the excavation.
The solid wall shaft struttings which have been described are
discontinued at the point where the shaft intersects the tunnel
excavation; and from this point to the floor of the tunnel an open
timbering is employed, whose only duty is to support the weight of
the solid strutting above. This timbering is made in various forms,
but the most common is a timber truss or arch construction which
spans the tunnel section.
Quantity of Timber.—The quantity of timber employed in
strutting a tunnel varies with the character of the material through
which the tunnel is excavated: it is small for solid-rock tunnels, and
large for soft-ground tunnels. In the Belgian method of excavation a
smaller quantity of timber is used than in any of the other ordinary
methods. For single-track tunnels excavated by this method there
will be needed on an average about 3 to 31⁄3 cu. yds. of timber per
lineal foot of tunnel. Practical experience shows that about four-fifths
of the timber once used can be employed for the second time. In
any of the methods in which the whole tunnel section is excavated
at once, the average amount of timber required per lineal foot is
about 8.7 cu. yds. Of this amount about two-thirds can be used a
second time. In the Italian method, in which the upper half and the
lower half are excavated separately, about 5 cu. yds. of timber are
required per lineal foot of tunnel, about one-half of which can be
employed a second time. For quicksand tunnels the amount of
timbering required per lineal foot varies from 3 to 5 cubic yds. Shaft
strutting requires from 1 to 11⁄2 cu. yds. of timber per lineal foot.
Dimensions of Timber.—The dimensions of the principal
members composing the strutting of headings, full section, and
shafts, are given in Table I. The planks used for lagging or the
poling-boards are usually from 4 ins. to 6 ins. wide, with a length
depending upon the method of strutting employed.

TABLE I.
Showing Sizes of Various Timbers Used in Strutting Tunnels Driven
Through Different Materials.

Rock. Soft Soils.

Com- Very
Hard. Soft. Loose.
pact. loose.
ins. ins. ins. ins. ins.
Headings:
Cap-pieces and vertical struts 6 8 10 12 14
Sills 8 10 12
Struts 5 5 6 7 8
Distance apart of the frames in feet 6 4.5 3 2.6 2.6
Strutting of the tunnel, longitudinal
strutting:
Crown bars 12 14 14
Props vertical or inclined supporting the
crown bars 10 12 14
 Sills 8 8 10
Cap-pieces or saddles 10 12 14
Struts to stiffen the structure 6 8 10
Distance apart of the frames (in feet) 4.5 4 3
Polygonal strutting:
Cap-pieces and contour pieces 8 10 12 14 16
Vertical struts on top 10 12 14 16 18
Vertical struts below 12 14 16 20 24
Intermediate sills 12 14 16 20 24
Lower sills 12 16 18
Raking props 10 10 10 12 12
Distance apart of the frames (in feet) 6 4.5 4 3 3
Shafts:
Horizontal beams forming the frame 8 8 10 12 14
Transverse beams 8 8 8 10 12
Vertical struts between the frames 8 8 10 12 12
Struts to reënforce the frame 6 8 8 8
Distance apart of the strutting (in feet) 6 4.5 4 3 2.6

IRON STRUTTING.

In 1862 Mr. Rziha employed old iron railway rails for strutting the
Naensen tunnel, and his example was successfully followed in
several tunnels built later where timber was scarce and expensive.
The advantages which iron strutting is claimed to possess over the
more common wooden structure are: its greater strength; the
smaller amount of space which it takes up; and the fact that it does
not wear out, and may, therefore, be used over and over again.
 Fig. 28.— Fig. 29.—Strutting
Strutting of made entirely of
Timber Posts Railway Rails.
and Railway
Rail Caps.

Iron Strutting in Headings.—In strutting the headings the

cross frames have a crown bar consisting of a section of old railway
rail carried either by wood or iron side posts. When wooden side
posts are used their upper ends have a dovetail mortise, and are
bound with an iron band, as shown by Fig. 28. The base of the rail
crown bar is set into the dovetail mortise and fastened by wedges.
When iron side posts are employed they usually consist of sections
of railway rails, and the crown bar is attached to them by fish-plate
connections, as shown by Fig. 29. The iron cross frames are set up
as the heading advances, and carry the plank lagging or poling-
boards, exactly in the same manner as the timber cross frames
previously described.
 Fig. 30.—Rziha’s Combined Strutting and Centering of Cast Iron.

Fig. 31.—Cast-Iron Segment of Rziha’s

Strutting and Centering.

Full Section Iron Strutting.—The iron strutting devised by Mr.

Rziha for full section work is shown by Fig. 30. Briefly described, it
consists of voussoir-shaped cast-iron segments, which are built up in
arch form. Fig. 31 shows the construction of one of the segments, all
of which are alike, with the exception of the crown segment, which
has a mortise and tenon joint which is kept open by filling the
mortise with sand. The segments are bolted together by means of
suitable bolt-holes in the vertical flanges, and when fully connected
form an arch rib of cast iron. This arch rib, A, Fig. 30, carries a series
of angle or T-iron frames bent into approximately voussoir shape, as
shown at B, Fig. 30. Above these frames are inserted the poling-
boards, running longitudinally, and spanning the distance between
consecutive arch ribs. By removing the bent iron frames the cast-iron
rib forms a center upon which to construct the masonry. Finally, to
remove the cast-iron rib itself, the sand is drawn out of the mortise
and tenon joint in the crown segment, which allows the joint to
close, and loosen the segments so that they are easily unbutted.
The illustration, Fig. 30, shows longitudinal poling-boards; more
often longitudinal crown bars of railway rails span the space between
connective arch ribs, and support transverse poling-boards. In
building the masonry, work is begun at the bottom on each side, the
bent iron frames being removed one after another to give room for
the masonry. As each frame is removed, it is replaced with a sort of
screw-jack to support the poling-boards until the masonry is
sufficiently completed to allow their removal. The interior bracing of
the arch rib shown at a a and b b consists of railway rails carried by
brackets cast on to the segments. A similar bracing of rails connects
the successive arch ribs. These lines of bracing serve to carry the
scaffolding upon which the masons work in building the lining.

Fig. 32.—Cast-Iron Segmental Strutting for Shafts.

Iron Shaft Strutting.—In soft-ground shaft work, the use of an

iron strutting, consisting of consecutive cast-iron rings, has
sometimes been employed to advantage. Fig. 32 shows the
construction of one of these rings, which, it will be seen, is
composed of four segments connected to each other by means of
bolted flanges. The holes shown in the circumferential web of the
ring are to allow for the seepage from the earth side walls. The
method of placing this cylindrical strutting is to start with a ring
having a cutting-edge. By means of excavation inside the ring, and
by ramming, the ring is sunk into the ground a distance equal to its
height. Another ring is then fastened by special hooks on top of the
first one, and the sinking continued until the second ring is down
flush with the surface. A third ring is then added, and so on until the
entire shaft is excavated and strutted. As in timber shaft strutting,
the solid iron ring strutting is carried down only to the top of the
tunnel section, and below this point there is an open timber or iron
supporting framework.
 CHAPTER VI.

METHODS OF HAULING IN TUNNELS.

The transportation from one point to another within the tunnel

and its shafts of any material, whether it is excavated spoil or
construction material, is defined as hauling. In all engineering
construction, the transportation of excavated materials, and
materials for construction, constitutes a very important part of the
expense of the work; but hauling in tunnels where the room is very
limited, and where work is constantly in progress over and at the
sides of the track, is a particularly expensive process. Hauling in
tunnels may be done either by way of the entrances, or by way of
the shafts, or by way of both the entrances and shafts.
 Fig. 33.—Platform Car for Tunnel Work.

Hauling by Way of Entrances.—When the hauling is done by

the way of the entrances, the materials to be hauled are taken
directly from the point of construction to the entrances, or in the
opposite direction, by means of special cars of different patterns. For
general purposes, these different patterns of cars may be grouped
into three classes,—platform-cars, dump-cars, and box-cars.
Representative examples of these several classes of cars are shown
in Figs. 33 to 36[6] inclusive, but it will be readily understood that
there are many other forms.
[6] Reproduced from catalogue of Arthur Koppel, New York.
Briefly described, platform-cars (Fig. 33) consist of a wooden
platform mounted on tracks, and they are usually employed for the
transportation of timber, ties, etc. Dump-cars are used in greater
numbers in tunnel work than any other form. Fig. 34 shows a dump-
car of metal construction, and Fig. 35 one constructed with a metal
under-frame and wooden box. These cars are made to run on
narrow-gauge tracks, and usually have a capacity of about one to
one and one-half cubic yards. Box-cars are more extensively
employed in Europe for tunnel work than in America. Fig. 36 shows
a typical European box-car for tunnel work. It is made either to run
on narrow-gauge or standard-gauge tracks.

Fig. 34.—Iron Dump-Car for Fig. 35.—Wooden Dump-Car

Tunnel Work. for Tunnel Work.

Fig. 36.—Box-Car for Tunnel Work.

 It is usually desirable in tunnel work to employ cars of different
forms for different parts of the work. In rock tunnels it is a common
practice to use narrow-gauge cars of small size in the headings, and
larger, broad-gauge cars for the enlargement of the profile. Where
narrow-gauge cars are employed for all purposes, it will also be
found more convenient to use platform-cars for handling the
construction material, and dump-cars for removing the spoil. The
extent to which it is desirable to use cars of different forms will
depend upon the character and conditions of the work, and
particularly upon how far it is possible to install the permanent track.
As a general ride, it is considered preferable to lay the permanent
tracks at once, and do all the hauling upon them, so that as soon as
the tunnel is completed, trains may pass through without delay. To
what extent this may be done, or whether it can be done at all or
not, depends upon the method of excavation and other local
conditions. In soft-ground tunnels excavated by the English or
Austrian methods, it is quite possible to lay the permanent tracks at
first, since the whole section is excavated at once, and the
excavation is kept but a little ahead of the completed tunnel. In rock
tunnels, where the heading is driven far ahead of the completed
section, it is, of course, impossible to keep the permanent track
close to the advance work, and narrow-gauge tracks must be laid in
the heading. The same thing is true in soft-ground tunnels driven by
successive headings and drifts. In these cases, therefore, where
narrow-gauge tracks have to be used for some portions of the work
anyway, the question comes up whether it is preferable to use
temporary narrow-gauge tracks throughout, or to lay the permanent
track as far ahead as possible, and then extend narrow-gauge tracks
to the advance excavation. In the latter case it will, of course, be
necessary to trans-ship each load from the narrow-gauge to the
standard-gauge cars, or vice versa, which means extra cost and
trouble. To avoid this, the method is sometimes adopted of laying a
third rail between the standard-gauge rails, so that either standard-
or narrow-gauge cars may be transported over the line. Whatever
form the local conditions may require the system of construction
tracks to assume, it may be set down as a general rule that the
permanent tracks should be kept as far advanced as possible, and
temporary tracks employed only where the permanent tracks are
impracticable.
The motive power employed for hauling in tunnels may be
furnished by animals or by mechanical motors. Animal power is
generally employed in short tunnels and in the advance headings
and galleries. In long tunnels, or where the excavated material has
to be transported some distance away from the tunnel, mechanical
power is preferable, for obvious reasons. The motors most used are
small steam locomotives, special compressed-air locomotives, and
electric motors. Compressed air and electric locomotives are built in
various forms, and are particularly well adapted for tunnel work
because of their small dimensions, and freedom from smoke and
heat.
Hauling by Way of Shafts.—When the excavated material and
materials of construction are handled through shafts, the operation
of hauling may be divided into three processes: the transportation of
the materials along the floor of the tunnel, the hoisting of them
through the shaft, and the surface transportation from and to the
mouth of the shaft. These three operations should be arranged to
work in harmony with each other, so as to avoid waste of time and
unnecessary handling of the materials. An endeavor should be made
to avoid, if possible, breaking or trans-shipping the load from the
time it starts at the heading until it is dumped at the spoil bank. This
can be accomplished in two ways. One way is to hoist the boxes of
the cars from their trucks at the bottom of the shaft, and place them
on similar trucks running on the surface tracks. The other way is to
run the loaded cars on to the elevator platform at the bottom, hoist
them, and then run them on to the surface tracks. If the first
method is employed, the car box is usually made of metal, and is
provided at its top edges with hooks or ears to which to attach the
hoisting cables. When the second method is used, the elevator
platform has tracks laid on it which connect with the tracks on the
tunnel floor, and also with those on the surface.
Hoisting Machinery.—The machines most commonly employed
for hoisting purposes in tunnel shafts are steam hoisting engines,
horse gins, and windlasses operated by hand. Windlasses and horse
gins are rather crude machines for hoisting loads, and are used only
in special circumstances, where the shaft is of small depth, when the
amount of material to be hoisted is small, or where for any reason
the use of hoisting engines is precluded. The steam hoisting engine
is the standard machine for the rapid lifting of heavy vertical loads.
Recently oil engines and electric hoists have also come to be used to
some extent, and under certain conditions these machines possess
notable advantages.
The construction of hand windlasses is familiar to every one. In
tunnel work this device is located directly over the shaft, with its axis
a little more than half a man’s height, so that the crank handle does
not rise above the shoulder line. To develop its greatest efficiency
the hoisting rope is passed around the windlass drum so that the
two ends hang down the shaft, and as one end descends the other
ascends. A skip, or bucket, is attached to each of the rope ends; and
by loading the descending skip with construction materials and the
ascending skip with spoil, the two skip loads tend to balance each
other, thus increasing the capacity of the windlass, and decreasing
the manual labor required to operate it. Skips varying from 0.3 cu.
yd. to 0.5 cu. yd. are used. The horse gin consists of a vertical
cylinder or drum provided with radial arms to which the horses are
hitched, which revolve the cylinder by walking around it in a circle.
The hoisting rope is rove around the drum so that the two ends
extend down the shaft with skips attached, as described in speaking
of the hand windlass. The power of the horse gin is, of course, much
greater than that of a windlass operated by hand, skips of 1 cu. yd.
capacity being commonly used. Horse gins are no longer economical
hoisting machines, according to one prominent authority, when
V(H + 20) > 5000, where V equals the volume of material to be
hoisted, and H equals the height of the hoist, the weight of the
excavated material being 2100 lbs. per cu. yd. As a general rule,
however, it is assumed that it is not economical to employ horse gins
with a depth of shaft exceeding 150 ft.
As already stated, the most efficient and most commonly used
device for hoisting at tunnel shafts is the steam hoisting engine.
There are numerous builders of hoisting engines, each of which
manufactures several patterns and sizes of engines. In each case,
however, the apparatus consists of a boiler supplying steam to a
horizontal engine which operates one or more rope drums. The
engines are always reversible. They may be employed to hoist the
skips directly, or to operate elevators upon which the skips or cars
are loaded. In either case the hoisting ropes pass from the engine
drum to and around vertical sheaves situated directly over the shaft
so as to secure the necessary vertical travel of the ropes down the
shaft. Where the shaft is divided into two compartments, each
having an elevator or hoist, double-drum engines are employed, one
drum being used for the operations in one compartment, and the
other for the operations in the other compartment. Where the work
is to be of considerable duration, or when it is done in cold weather,
more or less elaborate shelters or engine houses are built to cover
and protect the machinery.
Choice between the method of hoisting the skips directly, and the
method of using elevators, depends upon the extent and character
of the work. Where large quantities of material are to be hoisted
rapidly, it is generally considered preferable to employ elevators
instead of hoisting the skips directly. In direct hoisting at high speed,
oscillations are likely to be produced which may dash the skips
against the sides of the shaft and cause accidents. The loads which
can be carried in single skips are also smaller than those possible
where elevators are used; and this, combined with the slower
hoisting speed required, reduces the capacity of this method, as
compared with the use of elevators. Where elevators are employed,
however, the plant required is much more extensive and costly; it
comprising not only the elevator cars with their safety devices, etc.,
but the construction of a guiding framework for these cars in the
tunnel shaft. For these various reasons the elevator becomes the
preferable hoisting device where the quantity of material to be
handled is large, where the shafts are deep, and where the work will
extend over a long period of time; but when the contrary conditions
are the case, direct hoisting of the skips is generally the cheaper.
The engineer has to integrate the various factors in each individual
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com