handbook of computer vision and applications volume 2 signal processing and pattern recognition - - bernd jahne

HANDBOOK OF
COMPUTER VISION AND
APPLICATIONS
Volume 2
Signal Processing and
Pattern Recognition
ACADEMIC
PRESS
Bernd Jähne
Horst Haußecker
Peter Geißler
1
1
2
2
2
24
11
Handbook of
Computer Vision
and Applications
Volume 2
Signal Processing and
Pattern Recognition

Handbook of
Computer Vision
and Applications
Volume 2
Signal Processing and
Pattern Recognition

Editors
Bernd Jähne
Interdisciplinary Center for Scientific Computing
University of Heidelberg, Heidelberg, Germany
and
Scripps Institution of Oceanography
University of California, San Diego
Horst Haußecker
Peter Geißler
Interdisciplinary Center for Scientific Computing
University of Heidelberg, Heidelberg, Germany
ACADEMIC PRESS
San Diego London Boston
New York Sydney Tokyo Toronto
This book is printed on acid-free paper.
Copyright © 1999 by Academic Press.
All rights reserved.
No part of this publication may be reproduced or transmitted in any form or
by any means, electronic or mechanical, including photocopy, recording, or
any information storage and retrieval system, without permission in writing
from the publisher.
The appearance of code at the bottom of the first page of a chapter in this book
indicates the Publisher’s consent that copies of the chapter may be made for
personal or internal use of specific clients. This consent is given on the con-
dition, however, that the copier pay the stated per-copy fee through the Copy-
right Clearance Center, Inc. (222 Rosewood Drive, Danvers, Massachusetts
01923), for copying beyond that permitted by Sections 107 or 108 of the U.S.
Copyright Law. This consent does not extend to other kinds of copying, such
as copying for general distribution, for advertising or promotional purposes,
for creating new collective works, or for resale. Copy fees for pre-1999 chap-

ters are as shown on the title pages; if no fee code appears on the title page,
the copy fee is the same as for current chapters. ISBN 0-12-379770-5/$30.00
ACADEMIC PRESS
A Division of Harcourt Brace & Company
525 B Street, Suite 1900, San Diego, CA 92101-4495

ACADEMIC PRESS
24-28 Oval Road, London NW1 7DX, UK
/>Library of Congress Cataloging-In-Publication Data
Handbook of computer vision and applications / edited by Bernd Jähne,
Horst Haussecker, Peter Geissler.
p. cm.
Includes bibliographical references and indexes.
Contents: v. 1. Sensors and imaging—v. 2. Signal processing and
pattern recognition—v. 3. Systems and applications.
ISBN 0–12–379770–5 (set). — ISBN 0–12–379771-3 (v. 1)
ISBN 0–12–379772–1 (v. 2). — ISBN 0–12–379773-X (v. 3)
1. Computer vision — Handbooks, manuals. etc. I. Jähne, Bernd
1953– . II. Haussecker, Horst, 1968– . III. Geissler, Peter, 1966– .
TA1634.H36 1999
006.3

7 — dc21 98–42541
CIP
Printed in the United States of America
9900010203DS987654321
Contents
Preface xi
Contributors xiii
1 Introduction 1

B. Jähne
1.1 Signal processing for computer vision 2
1.2 Pattern recognition for computer vision 3
1.3 Computational complexity and fast algorithms 4
1.4 Performance evaluation of algorithms 5
1.5 References 6
I Signal Representation
2 Continuous and Digital Signals 9
B. Jähne
2.1 Introduction 10
2.2 Continuous signals 10
2.3 Discrete signals 13
2.4 Relation between continuous and discrete signals 23
2.5 Quantization 30
2.6 References 34
3 Spatial and Fourier Domain 35
B. Jähne
3.1 Vector spaces and unitary transforms 35
3.2 Continuous Fourier transform (FT) 41
3.3 The discrete Fourier transform (DFT) 51
3.4 Fast Fourier transform algorithms (FFT) 57
3.5 References 66
4 Multiresolutional Signal Representation 67
B. Jähne
4.1 Scale in signal processing 67
4.2 Scale filters 70
4.3 Scale space and diffusion 76
4.4 Multigrid representations 84
4.5 References 90
v

vi Contents
II Elementary Spatial Processing
5 Neighborhood Operators 93
B. Jähne
5.1 Introduction 94
5.2 Basics 94
5.3 Linear shift-invariant filters 98
5.4 Recursive filters 106
5.5 Classes of nonlinear filters 113
5.6 Efficient neighborhood operations 116
5.7 References 124
6 Principles of Filter Design 125
B. Jähne, H. Scharr, and S. Körkel
6.1 Introduction 125
6.2 Filter design criteria 126
6.3 Windowing techniques 128
6.4 Filter cascading 132
6.5 Filter design as an optimization problem 133
6.6 Design of steerable filters and filter families 143
6.7 References 151
7 Local Averaging 153
B. Jähne
7.1 Introduction 153
7.2 Basic features 154
7.3 Box filters 158
7.4 Binomial filters 163
7.5 Cascaded averaging 167
7.6 Weighted averaging 173
7.7 References 174
8 Interpolation 175

B. Jähne
8.1 Introduction 175
8.2 Basics 176
8.3 Interpolation in Fourier space 180
8.4 Polynomial interpolation 182
8.5 Spline-based interpolation 187
8.6 Optimized interpolation 190
8.7 References 192
9 Image Warping 193
B. Jähne
9.1 Introduction 193
9.2 Forward and inverse mapping 194
9.3 Basic geometric transforms 195
9.4 Fast algorithms for geometric transforms 199
9.5 References 206
Contents vii
III Feature Estimation
10 Local Structure 209
B. Jähne
10.1 Introduction 210
10.2 Properties of simple neighborhoods 210
10.3 Edge detection by first-order derivatives 213
10.4 Edge detection by zero crossings 223
10.5 Edges in multichannel images 226
10.6 First-order tensor representation 227
10.7 References 238
11 Principles for Automatic Scale Selection 239
T. Lindeberg
11.1 Introduction 240
11.2 Multiscale differential image geometry 240

11.3 A general scale-selection principle 247
11.4 Feature detection with automatic scale selection 251
11.5 Feature localization with automatic scale selection 262
11.6 Stereo matching with automatic scale selection 265
11.7 Summary and conclusions 269
11.8 References 270
12 Texture Analysis 275
T. Wagner
12.1 Importance of texture 276
12.2 Feature sets for texture analysis 278
12.3 Assessment of textural features 299
12.4 Automatic design of texture analysis systems 306
12.5 References 307
13 Motion 309
H. Haußecker and H. Spies
13.1 Introduction 310
13.2 Basics: flow and correspondence 312
13.3 Optical flow-based motion estimation 321
13.4 Quadrature filter techniques 345
13.5 Correlation and matching 353
13.6 Modeling of flow fields 356
13.7 Confidence measures and error propagation 369
13.8 Comparative analysis 373
13.9 References 392
14 Bayesian Multiscale Differential Optical Flow 397
E. P. Simoncelli
14.1 Introduction 397
14.2 Differential formulation 398
14.3 Uncertainty model 400
14.4 Coarse-to-fine estimation 404

14.5 Implementation issues 410
14.6 Examples 414
14.7 Conclusion 419
14.8 References 420
viii Contents
15 Nonlinear Diffusion Filtering 423
J. Weickert
15.1 Introduction 424
15.2 Filter design 425
15.3 Continuous theory 433
15.4 Algorithmic details 436
15.5 Discrete theory 439
15.6 Parameter selection 441
15.7 Generalizations 444
15.8 Summary 446
15.9 References 446
16 Variational Methods 451
C. Schnörr
16.1 Introduction 451
16.2 Processing of two- and three-dimensional images 455
16.3 Processing of vector-valued images 471
16.4 Processing of image sequences 476
16.5 References 481
17 Stereopsis - Geometrical and Global Aspects 485
H. A. Mallot
17.1 Introduction 485
17.2 Stereo geometry 487
17.3 Global stereopsis 499
17.4 References 502
18 Stereo Terrain Reconstruction by Dynamic Programming 505

G. Gimel’farb
18.1 Introduction 505
18.2 Statistical decisions in terrain reconstruction 509
18.3 Probability models of epipolar profiles 514
18.4 Dynamic programming reconstruction 520
18.5 Experimental results 524
18.6 References 528
19 Reflectance-Based Shape Recovery 531
R. Klette, R. Kozera, and K. Schlüns
19.1 Introduction 532
19.2 Reflection and gradients 539
19.3 Three light sources 552
19.4 Two light sources 559
19.5 Theoretical framework for shape from shading 571
19.6 Shape from shading 574
19.7 Concluding remarks 586
19.8 References 587
20 Depth-from-Focus 591
P. Geißler and T. Dierig
20.1 Introduction 592
20.2 Basic concepts 593
20.3 Principles of depth-from-focus algorithms 595
Contents ix
20.4 Multiple-view depth-from-focus 596
20.5 Dual-view depth-from-focus 601
20.6 Single-view depth-from-focus 608
20.7 References 622
IV Object Analysis, Classification, Modeling, Visualization
21 Morphological Operators 627
P. Soille

21.1 Introduction 628
21.2 Basics 629
21.3 Morphological operators 637
21.4 Efficient computation of morphological operators 659
21.5 Morphological image processing 664
21.6 References 678
22 Fuzzy Image Processing 683
H. Haußecker and H. R. Tizhoosh
22.1 Introduction 684
22.2 Why fuzzy image processing? 691
22.3 Fuzzy image understanding 692
22.4 Fuzzy image processing systems 699
22.5 Theoretical components of fuzzy image processing 702
22.6 Selected application examples 714
22.7 Conclusions 721
22.8 References 722
23 Neural Net Computing for Image Processing 729
A. Meyer-Bäse
23.1 Introduction 729
23.2 Multilayer perceptron (MLP) 730
23.3 Self-organizing neural networks 736
23.4 Radial-basis neural networks (RBNN) 740
23.5 Transformation radial-basis networks (TRBNN) 743
23.6 Hopfield neural networks 747
23.7 References 751
24 Graph Theoretical Concepts for Computer Vision 753
D. Willersinn et al.
24.1 Introduction 754
24.2 Basic definitions 754
24.3 Graph representation of two-dimensional digital images 760

24.4 Voronoi diagrams and Delaunay graphs 762
24.5 Matching 775
24.6 Graph grammars 780
24.7 References 786
25 Shape Reconstruction from Volumetric Data 791
R. Eils and K. Sätzler
25.1 Introduction 791
25.2 Incremental approach 794
x Contents
25.3 Three-dimensional shape reconstruction from contour lines . 797
25.4 Volumetric shape reconstruction 802
25.5 Summary 811
25.6 References 813
26 Probabilistic Modeling in Computer Vision 817
J. Hornegger, D. Paulus, and H. Niemann
26.1 Introduction 817
26.2 Why probabilistic models? 819
26.3 Object recognition: classification and regression 821
26.4 Parametric families of model densities 826
26.5 Automatic model generation 844
26.6 Practical issues 850
26.7 Summary, conclusions, and discussion 852
26.8 References 852
27 Knowledge-Based Interpretation of Images 855
H. Niemann
27.1 Introduction 855
27.2 Model of the task domain 859
27.3 Interpretation by optimization 864
27.4 Control by graph search 865
27.5 Control by combinatorial optimization 868

27.6 Judgment function 870
27.7 Extensions and remarks 872
27.8 References 872
28 Visualization of Volume Data 875
J. Hesser and C. Poliwoda
28.1 Selected visualization techniques 876
28.2 Basic concepts and notation for visualization 880
28.3 Surface rendering algorithms and OpenGL 881
28.4 Volume rendering 884
28.5 The graphics library VGL 890
28.6 How to use volume rendering 898
28.7 Volume rendering 901
28.8 Acknowledgments 905
28.9 References 905
29 Databases for Microscopes and Microscopical Images 907
N. Salmon, S. Lindek, and E. H. K. Stelzer
29.1 Introduction 908
29.2 Towards a better system for information management 909
29.3 From flat files to database systems 911
29.4 Database structure and content 912
29.5 Database system requirements 917
29.6 Data flow—how it looks in practice 918
29.7 Future prospects 921
29.8 References 925
Index 927
Preface
What this handbook is about
This handbook offers a fresh approach to computer vision. The whole
vision process from image formation to measuring, recognition, or re-
acting is regarded as an integral process. Computer vision is under-

stood as the host of techniques to acquire, process, analyze, and un-
derstand complex higher-dimensional data from our environment for
scientific and technical exploration.
In this sense the handbook takes into account the interdisciplinary
nature of computer vision with its links to virtually all natural sciences
and attempts to bridge two important gaps. The first is between mod-
ern physical sciences and the many novel techniques to acquire images.
The second is between basic research and applications. When a reader
with a background in one of the fields related to computer vision feels
he has learned something from one of the many other facets of com-
puter vision, the handbook will have fulfilled its purpose.
The handbook comprises three volumes. The first volume, Sensors
and Imaging, covers image formation and acquisition. The second vol-
ume, Signal Processing and Pattern Recognition , focuses on processing
of the spatial and spatiotemporal signal acquired by imaging sensors.
The third volume, Systems and Applications, describes how computer
vision is integrated into systems and applications.
Prerequisites
It is assumed that the reader is familiar with elementary mathematical
concepts commonly used in computer vision and in many other areas
of natural sciences and technical disciplines. This includes the basics
of set theory, matrix algebra, differential and integral equations, com-
plex numbers, Fourier transform, probability, random variables, and
graphing. Wherever possible, mathematical topics are described intu-
itively. In this respect it is very helpful that complex mathematical
relations can often be visualized intuitively by images. For a more for-
xi
xii Preface
mal treatment of the corresponding subject including proofs, suitable
references are given.

How to use this handbook
The handbook has been designed to cover the different needs of its
readership. First, it is suitable for sequential reading. In this way the
reader gets an up-to-date account of the state of computer vision. It is
presented in a way that makes it accessible for readers with different
backgrounds. Second, the reader can look up specific topics of inter-
est. The individual chapters are written in a self-consistent way with
extensive cross-referencing to other chapters of the handbook and ex-
ternal references. The CD that accompanies each volume of the hand-
book contains the complete text of the handbook in the Adobe Acrobat
portable document file format (PDF). This format can be read on all
major platforms. Free Acrobat reader version 3.01 for all major com-
puting platforms is included on the CDs. The texts are hyperlinked in
multiple ways. Thus the reader can collect the information of interest
with ease. Third, the reader can delve more deeply into a subject with
the material on the CDs. They contain additional reference material,
interactive software components, code examples, image material, and
references to sources on the Internet. For more details see the readme
file on the CDs.
Acknowledgments
Writing a handbook on computer vision with this breadth of topics is
a major undertaking that can succeed only in a coordinated effort that
involves many co-workers. Thus the editors would like to thank first
all contributors who were willing to participate in this effort. Their
cooperation with the constrained time schedule made it possible that
the three-volume handbook could be published in such a short period
following the call for contributions in December 1997. The editors are
deeply grateful for the dedicated and professional work of the staff at
AEON Verlag & Studio who did most of the editorial work. We also
express our sincere thanks to Academic Press for the opportunity to

write this handbook and for all professional advice.
Last but not least, we encourage the reader to send us any hints
on errors, omissions, typing errors, or any other shortcomings of the
handbook. Actual information about the handbook can be found at the
editors homepage .
Heidelberg, Germany and La Jolla, California, December 1998
Bernd Jähne, Horst Haußecker, Peter Geißler
Contributors
Etienne Bertin received the PhD degree in mathematics
from Université Joseph Fourier in 1994. From 1990 to
1995 he worked on various topics in image analysis and
computational geometry. Since 1995, he has been an as-
sistant professor at the Université Pierre Mendès France
in the Laboratoire de statistique et d’analyses de don-
nées; he works on stochastic geometry.
Dr. Etienne Bertin
Laboratoire de Statistique et d’analyse de donnés
Université Pierre Mendès, Grenoble, France

Anke Meyer-Bäse received her M. S. and the PhD in elec-
trical engineering from the Darmstadt Institute of Tech-
nology in 1990 and 1995, respectively. From 1995 to
1996 she was a postdoctoral fellow with the Federal Insti-
tute of Neurobiology, Magdeburg, Germany. Since 1996
she was a visiting assistant professor with the Dept. of
Electrical Engineering, University of Florida, Gainesville,
USA. She received the Max-Kade award in Neuroengineer-
ing in 1996 and the Lise-Meitner prize in 1997. Her re-
search interests include neural networks, image process-
ing, biomedicine, speech recognition, and theory of non-

linear systems.
Dr. Anke Meyer-Bäse, Dept. of Electrical Engineering and Computer Science,
University of Florida, 454 New Engineering Building 33, Center Drive
PO Box 116130, Gainesville, FL 32611-6130, U.S., fl.edu
Tobias Dierig graduated in 1997 from the University of
Heidelberg with a master degree in physics and is now
pursuing his PhD at the Interdisciplinary Center for Sci-
entific Computing at Heidelberg university. He is con-
cerned mainly with depth from focus algorithms, image
fusion, and industrial applications of computer vision
within the OpenEye project.
Tobias Dierig, Forschungsgruppe Bildverarbeitung, IWR
Universität Heidelberg, Im Neuenheimer Feld 368
D-69120 Heidelberg, Germany

/>xiii
xiv Contributors
Roland Eils studied mathematics and computer science
in Aachen, where he received his diploma in 1990. After
a two year stay in Indonesia for language studies he joint
the Graduiertenkolleg “Modeling and Scientific Comput-
ing in Mathematics and Sciences” at the Interdisciplinary
Center for Scientific Computing (IWR), University of Hei-
delberg, where he received his doctoral degree in 1995.
Since 1996 he has been leading the biocomputing group,
Structures in Molecular Biology. His research interests
include computer vision, in particular computational ge-
ometry, and application of image processing techniques
in science and biotechnology.
Dr. Roland Eils, Biocomputing-Gruppe, IWR, Universität Heidelberg

Im Neuenheimer Feld 368, D-69120 Heidelberg, Germany

/>Peter Geißler studied physics in Heidelberg. He received
his diploma and doctoral degree from Heidelberg Uni-
versity in 1994 and 1998, respectively. His research in-
terests include computer vision, especially depth-from-
focus, adaptive filtering, and flow visualization as well as
the application of image processing in physical sciences
and oceanography.
Dr. Peter Geißler
Forschungsgruppe Bildverarbeitung, IWR
Universität Heidelberg, Im Neuenheimer Feld 368
D-69120 Heidelberg, Germany


Georgy Gimel’farb received his PhD degree from the
Ukrainian Academy of Sciences in 1969 and his Doctor of
Science (the habilitation) degree from the Higher Certify-
ing Commission of the USSR in 1991. In 1962, he began
working in the Pattern Recognition, Robotics, and Image
Recognition Departments of the Institute of Cybernetics
(Ukraine). In 1994–1997 he was an invited researcher in
Hungary, the USA, Germany, and France. Since 1997, he
has been a senior lecturer in computer vision and digital
TV at the University of Auckland, New Zealand. His re-
search interests include analysis of multiband space and
aerial images, computational stereo, and image texture analysis.
Dr. Georgy Gimel’farb, Centre for Image Technology and Robotics,
Department of Computer Science, Tamaki Campus
The University of Auckland, Private Bag 92019, Auckland 1, New Zealand

, />Contributors xv
Horst Haußecker studied physics in Heidelberg. He re-
ceived his diploma in physics and his doctoral degree
from Heidelberg University in 1994 and 1996, respec-
tively. He was visiting scientist at the Scripps Institution
of Oceanography in 1994. Currently he is conducting
research in the image processing research group at the
Interdisciplinary Center for Scientific Computing (IWR),
where he also lectures on optical flow computation. His
research interests include computer vision, especially
image sequence analysis, infrared thermography, and
fuzzy-image processing, as well as the application of im-
age processing in physical sciences and oceanography.
Dr. Horst Haußecker, Forschungsgruppe Bildverarbeitung, IWR
Universität Heidelberg, Im Neuenheimer Feld 368, D-69120 Heidelberg


Jürgen Hesser is assistant professor at the Lehrstuhl für
Informatik V, University of Mannheim, Germany. He
heads the groups on computer graphics, bioinformat-
ics, and optimization. His research interests are real-
time volume rendering, computer architectures, compu-
tational chemistry, and evolutionary algorithms. In addi-
tion, he is co-founder of Volume Graphics GmbH, Heidel-
berg. Hesser received his PhD and his diploma in physics
at the University of Heidelberg, Germany.
Jürgen Hesser, Lehrstuhl für Informatik V
Universität Mannheim
B6, 26, D-68131 Mannheim, Germany
,

Joachim Hornegger graduated in 1992 and received his
PhD degree in computer science in 1996 from the Uni-
versität Erlangen-Nürnberg, Germany, for his work on
statistical object recognition. Joachim Hornegger was
research and teaching associate at Universität Erlangen-
Nürnberg, a visiting scientist at the Technion, Israel, and
at the Massachusetts Institute of Technology, U.S. He
is currently a research scholar and teaching associate
at Stanford University, U.S. Joachim Hornegger is the
author of 30 technical papers in computer vision and
speech processing and three books. His research inter-
ests include 3-D computer vision, 3-D object recognition, and statistical meth-
ods applied to image analysis problems.
Dr. Joachim Hornegger, Stanford University, Robotics Laboratory
Gates Building 1A, Stanford, CA 94305-9010, U.S.
, />xvi Contributors
Bernd Jähne studied physics in Saarbrücken and Hei-
delberg. He received his diploma, doctoral degree, and
habilitation degree from Heidelberg University in 1977,
1980, and 1985, respectively, and a habilitation de-
gree in applied computer science from the University of
Hamburg-Harburg in 1992. Since 1988 he has been a Ma-
rine Research Physicist at Scripps Institution of Oceanog-
raphy, University of California, and, since 1994, he has
been professor of physics at the Interdisciplinary Center
of Scientific Computing. He leads the research group on
image processing. His research interests include com-
puter vision, especially filter design and image sequence
analysis, the application of image processing techniques
in science and industry, and small-scale air-sea interaction processes.

Prof. Dr. Bernd Jähne, Forschungsgruppe Bildverarbeitung, IWR
Universität Heidelberg, Im Neuenheimer Feld 368, D-69120 Heidelberg


Reinhard Klette studied mathematics at Halle University,
received his master degree and doctor of natural science
degree in mathematics at Jena University, became a do-
cent in computer science, and was a professor of com-
puter vision at Berlin Technical University. Since June
1996 he has been professor of information technology
in the Department of Computer Science at the University
of Auckland. His research interests include theoretical
and applied topics in image processing, pattern recogni-
tion, image analysis, and image understanding. He has
published books about image processing and shape reconstruction and was
chairman of several international conferences and workshops on computer
vision. Recently, his research interests have been directed at 3-D biomedical
image analysis with digital geometry and computational geometry as major
subjects.
Prof. Dr. Reinhard Klette, Centre for Image Technology and Robotics,
Computer Science Department, Tamaki Campus
The Auckland University, Private Bag 92019, Auckland, New Zealand
, />Christoph Klauck received his diploma in computer sci-
ence and mathematics from the University of Kaiser-
slautern, Germany, in 1990. From 1990 to 1994 he
worked as research scientist at the German Research
Center for Artificial Intelligence Inc. (DFKI GmbH) at
Kaiserslautern. In 1994 he finished his dissertation in
computer science. Since then he has been involved in
the IRIS project at the University of Bremen (Artificial

Intelligence Group). His primary research interests in-
clude graph grammars and rewriting systems in general,
knowledge representation, and ontologies.
Contributors xvii
Prof. Dr. Christoph Klauck, Dep. of Electrical Eng. and Computer Science
University of Hamburg (FH), Berliner Tor 3, D-20099 Hamburg, Germany
, />Stefan Körkel is member of the research groups for nu-
merics and optimization of Prof. Bock and Prof. Reinelt
at the Interdisciplinary Center for Scientific Computing
at the University of Heidelberg, Germany. He studied
mathematics in Heidelberg. Currently he is pursuing his
PhD in nonlinear and mixed integer optimization meth-
ods. His research interests include filter optimization as
well as nonlinear optimum experimental design.
Stefan Körkel
Interdisciplinary Center for Scientific Computing
Im Neuenheimer Feld 368, 69120 Heidelberg

/>Ryszard Kozera received his M.Sc. degree in pure mathe-
matics in 1985 from Warsaw University, Poland, his PhD
degree in computer science in 1991 from Flinders Uni-
versity, Australia, and finally his PhD degree in mathe-
matics in 1992 from Warsaw University, Poland. He is
currently employed as a senior lecturer at the University
of Western Australia. Between July 1995 and February
1997, Dr. Kozera was at the Technical University of Berlin
and at Warsaw University as an Alexander von Humboldt
Foundation research fellow. His current research inter-
ests include applied mathematics with special emphasis
on partial differential equations, computer vision, and

numerical analysis.
Dr. Ryszard Kozera, Department of Computer Science, The University of West-
ern Australia, Nedlands, WA 6907, Australia,
/>Tony Lindeberg received his M.Sc. degree in engineer-
ing physics and applied mathematics from KTH (Royal
Institute of Technology), Stockholm, Sweden in 1987,
and his PhD degree in computing science in 1991. He
is currently an associate professor at the Department
of Numerical Analysis and Computing Science at KTH.
His main research interests are in computer vision and
relate to multiscale representations, focus-of-attention,
and shape. He has contributed to the foundations of
continuous and discrete scale-space theory, as well as
to the application of these theories to computer vision
problems. Specifically, he has developed principles for
automatic scale selection, methodologies for extracting salient image struc-
tures, and theories for multiscale shape estimation. He is author of the book
“Scale-Space Theory in Computer Vision.”
xviii Contributors
Tony Lindeberg, Department of Numerical Analysis and Computing Science
KTH, S-100 44 Stockholm, Sweden.
, />Steffen Lindek studied physics at the RWTH Aachen, Ger-
many, the EPF Lausanne, Switzerland, and the Univer-
sity of Heidelberg, Germany. He did his diploma and
PhD theses in the Light Microscopy Group at the Euro-
pean Molecular Biology Laboratory (EMBL), Heidelberg,
Germany, developing high-resolution light-microscopy
techniques. Since December 1996 he has been a post-
doctoral fellow with the BioImage project at EMBL. He
currently works on the design and implementation of the

image database, and he is responsible for the administra-
tion of EMBL’s contribution to the project.
Dr. Steffen Lindek, European Molecular Biology Laboratory (EMBL)
Postfach 10 22 09, D-69120 Heidelberg, Germany

Hanspeter A. Mallot studied biology and mathematics at
the University of Mainz where he also received his doc-
toral degree in 1986. He was a postdoctoral fellow at
the Massachusetts Institute of Technology in 1986/87
and held research positions at Mainz University and the
Ruhr-Universität-Bochum. In 1993, he joined the Max-
Planck-Institut für biologische Kybernetik in Tübingen.
In 1996/97, he was a fellow at the Institute of Advanced
Studies in Berlin. His research interests include the per-
ception of shape and space in humans and machines,
cognitive maps, as well as neural network models of the
cerebral cortex.
Dr. Hanspeter A. Mallot, Max-Planck-Institut für biologische Kybernetik
Spemannstr. 38, 72076 Tübingen, Germany

/>Heinrich Niemann obtained the degree of Dipl Ing. in
electrical engineering and Dr Ing. at Technical Univer-
sity Hannover in 1966 and 1969, respectively. From
1967 to 1972 he was with Fraunhofer Institut für In-
formationsverarbeitung in Technik und Biologie, Karls-
ruhe. Since 1975 he has been professor of computer sci-
ence at the University of Erlangen-Nürnberg and since
1988 he has also served as head of the research group,
Knowledge Processing, at the Bavarian Research Institute
for Knowledge-Based Systems (FORWISS). His fields of

research are speech and image understanding and the
application of artificial intelligence techniques in these
fields. He is the author or co-author of 6 books and approximately 250 jour-
nal and conference contributions.
Contributors xix
Prof. Dr Ing. H. Niemann, Lehrstuhl für Mustererkennung (Informatik 5)
Universität Erlangen-Nürnberg, Martensstraße 3, 91058 Erlangen, Germany


Dietrich Paulus received a bachelor degree in computer
science at the University of Western Ontario, London,
Canada (1983). He graduated (1987) and received his
PhD degree (1991) from the University of Erlangen-
Nürnberg, Germany. He is currently a senior researcher
(Akademischer Rat) in the field of image pattern recog-
nition and teaches courses in computer vision and ap-
plied programming for image processing. Together with
J. Hornegger, he has recently written a book on pattern
recognition and image processing in C++.
Dr. Dietrich Paulus, Lehrstuhl für Mustererkennung
Universität Erlangen-Nürnberg, Martensstr. 3, 91058 Erlangen, Germany


Christoph Poliwoda is PhD student at the Lehrstuhl für
Informatik V, University of Mannheim, and leader of the
development section of Volume Graphics GmbH. His re-
search interests are real-time volume and polygon ray-
tracing, 3-D image processing, 3-D segmentation, com-
puter architectures and parallel computing. Poliwoda
received his diploma in physics at the University of Hei-

delberg, Germany.
Christoph Poliwoda
Lehrstuhl für Informatik V
Universität Mannheim
B6, 26, D-68131 Mannheim, Germany

Nicholas J. Salmon received the master of engineering
degree from the Department of Electrical and Electronic
Engineering at Bath University, England, in 1990. Then
he worked as a software development engineer for Mar-
coni Radar Systems Ltd., England, helping to create a
vastly parallel signal-processing machine for radar appli-
cations. Since 1992 he has worked as software engineer
in the Light Microscopy Group at the European Molecu-
lar Biology Laboratory, Germany, where he is concerned
with creating innovative software systems for the con-
trol of confocal microscopes, and image processing.
Nicholas J. Salmon, Light Microscopy Group,
European Molecular Biology Laboratory (EMBL)
Postfach 10 22 09, D-69120 Heidelberg, Germany
,
xx Contributors
Kurt Sätzler studied physics at the University of Hei-
delberg, where he received his diploma in 1995. Since
then he has been working as a PhD student at the Max-
Planck-Institute of Medical Research in Heidelberg. His
research interests are mainly computational geometry
applied to problems in biomedicine, architecture and
computer graphics, image processing and tilted view mi-
croscopy.

Kurt Sätzler, IWR, Universität Heidelberg
Im Neuenheimer Feld 368, D-69120 Heidelberg
or
Max-Planck-Institute for Medical Research, Department of Cell Physiology
Jahnstr. 29, D-69120 Heidelberg, Germany

Hanno Scharr studied physics at the University of Hei-
delberg, Germany and did his diploma thesis on tex-
ture analysis at the Interdisciplinary Center for Scien-
tific Computing in Heidelberg. Currently, he is pursu-
ing his PhD on motion estimation. His research interests
include filter optimization and motion estimation in dis-
crete time series of n-D images.
Hanno Scharr
Interdisciplinary Center for Scientific Computing
Im Neuenheimer Feld 368, 69120 Heidelberg, Germany

/>Karsten Schlüns studied computer science in Berlin. He
received his diploma and doctoral degree from the Tech-
nical University of Berlin in 1991 and 1996. From 1991 to
1996 he was research assistant in the Computer Vision
Group, Technical University of Berlin, and from 1997
to 1998 he was a postdoctoral research fellow in com-
puting and information technology, University of Auck-
land. Since 1998 he has been a scientist in the image
processing group at the Institute of Pathology, Univer-
sity Hospital Charité in Berlin. His research interests
include pattern recognition and computer vision, espe-
cially three-dimensional shape recovery, performance analysis of reconstruc-
tion algorithms, and teaching of computer vision.

Dr. Karsten Schlüns, Institute of Pathology,
University Hospital Charité, Schumannstr. 20/21, D-10098 Berlin, Germany
, />Contributors xxi
Christoph Schnörr received the master degree in electri-
cal engineering in 1987, the doctoral degree in computer
science in 1991, both from the University of Karlsruhe
(TH), and the habilitation degree in Computer Science in
1998 from the University of Hamburg, Germany. From
1987–1992, he worked at the Fraunhofer Institute for In-
formation and Data Processing (IITB) in Karlsruhe in the
field of image sequence analysis. In 1992 he joined the
Cognitive Systems group, Department of Computer Sci-
ence, University of Hamburg, where he became an assis-
tant professor in 1995. He received an award for his work
on image segmentation from the German Association for
Pattern Recognition (DAGM) in 1996. Since October 1998, he has been a full
professor at the University of Mannheim, Germany, where he heads the Com-
puter Vision, Graphics, and Pattern Recognition Group. His research interests
include pattern recognition, machine vision, and related aspects of computer
graphics, machine learning, and applied mathematics.
Prof. Dr. Christoph Schnörr, University of Mannheim
Dept. of Math. & Computer Science, D-68131 Mannheim, Germany
,
Eero Simoncelli started his higher education with a bach-
elor’s degree in physics from Harvard University, went
to Cambridge University on a fellowship to study mathe-
matics for a year and a half, and then returned to the USA
to pursue a doctorate in Electrical Engineering and Com-
puter Science at MIT. He received his PhD in 1993, and
joined the faculty of the Computer and Information Sci-

ence Department at the University of Pennsylvania that
same year. In September of 1996, he joined the faculty
of the Center for Neural Science and the Courant Insti-
tute of Mathematical Sciences at New York University. He
received an NSF Faculty Early Career Development (CA-
REER) grant in September 1996, for teaching and research in “Visual Informa-
tion Processing”, and a Sloan Research Fellowship in February 1998.
Dr. Eero Simoncelli, 4 Washington Place, RM 809, New York, NY 10003-6603
, />Pierre Soille received the engineering degree from the
Université catholique de Louvain, Belgium, in 1988. He
gained the doctorate degree in 1992 at the same univer-
sity and in collaboration with the Centre de Morphologie
Mathématique of the Ecole des Mines de Paris. He then
pursued research on image analysis at the CSIRO Math-
ematical and Information Sciences Division, Sydney, the
Centre de Morphologie Mathématique of the Ecole des
Mines de Paris, and the Abteilung Mustererkennung of
the Fraunhofer-Institut IPK, Berlin. During the period
1995-1998 he was lecturer and research scientist at the
Ecole des Mines d’Alès and EERIE, Nîmes, France. Now he is a senior research
scientist at the Silsoe Research Institute, England. He worked on many ap-
xxii Contributors
plied projects, taught tutorials during international conferences, co-organized
the second International Symposium on Mathematical Morphology, wrote and
edited three books, and contributed to over 50 scientific publications.
Prof. Pierre Soille, Silsoe Research Institute, Wrest Park
Silsoe, Bedfordshire, MK45 4HS, United Kingdom
,
Hagen Spies graduated in January 1998 from the Univer-
sity of Heidelberg with a master degree in physics. He

also received an MS in computing and information tech-
nology from the University of Dundee, Scotland in 1995.
In 1998/1999 he spent one year as a visiting scientist at
the University of Western Ontario, Canada. Currently he
works as a researcher at the Interdisciplinary Center for
Scientific Computing at the University of Heidelberg. His
interests concern the measurement of optical and range
flow and their use in scientific applications.
Hagen Spies, Forschungsgruppe Bildverarbeitung, IWR
Universität Heidelberg, Im Neuenheimer Feld 368
D-69120 Heidelberg, Germany,
/>E. H. K. Stelzer studied physics in Frankfurt am Main and
in Heidelberg, Germany. During his Diploma thesis at
the Max-Planck-Institut für Biophysik he worked on the
physical chemistry of phospholipid vesicles, which he
characterized by photon correlation spectroscopy. Since
1983 he has worked at the European Molecular Biol-
ogy Laboratory (EMBL). He has contributed extensively
to the development of confocal fluorescence microscopy
and its application in life sciences. His group works
on the development and application of high-resolution
techniques in light microscopy, video microscopy, con-
focal microscopy, optical tweezers, single particle analy-
sis, and the documentation of relevant parameters with biological data.
Prof. Dr. E. H. K. Stelzer, Light Microscopy Group,
European Molecular Biology Laboratory (EMBL), Postfach 10 22 09
D-69120 Heidelberg, Germany, ,
Hamid R. Tizhoosh received the M.S. degree in electrical
engineering from University of Technology, Aachen, Ger-
many, in 1995. From 1993 to 1996, he worked at Man-

agement of Intelligent Technologies Ltd. (MIT GmbH),
Aachen, Germany, in the area of industrial image pro-
cessing. He is currently a PhD candidate, Dept. of Tech-
nical Computer Science of Otto-von-Guericke-University,
Magdeburg, Germany. His research encompasses fuzzy
logic and computer vision. His recent research efforts
include medical and fuzzy image processing. He is cur-
rently involved in the European Union project INFOCUS,
and is researching enhancement of medical images in radiation therapy.
H. R. Tizhoosh, University of Magdeburg (IPE)
Contributors xxiii
P.O. Box 4120, D-39016 Magdeburg, Germany

/>Thomas Wagner received a diploma degree in physics in
1991 from the University of Erlangen, Germany. In 1995,
he finished his PhD in computer science with an applied
image processing topic at the Fraunhofer Institute for In-
tegrated Circuits in Erlangen. Since 1992, Dr. Wagner has
been working on industrial image processing problems
at the Fraunhofer Institute, from 1994 to 1997 as group
manager of the intelligent systems group. Projects in
his research team belong to the fields of object recogni-
tion, surface inspection, and access control. In 1996, he
received the “Hans-Zehetmair-Habilitationsförderpreis.”
He is now working on automatic solutions for the design
of industrial image processing systems.
Dr Ing. Thomas Wagner, Fraunhofer Institut für Intregrierte Schaltungen
Am Weichselgarten 3, D-91058 Erlangen, Germany
,
Joachim Weickert obtained a M.Sc. in industrial math-

ematics in 1991 and a PhD in mathematics in 1996,
both from Kaiserslautern University, Germany. After re-
ceiving the PhD degree, he worked as post-doctoral re-
searcher at the Image Sciences Institute of Utrecht Uni-
versity, The Netherlands. In April 1997 he joined the
computer vision group of the Department of Computer
Science at Copenhagen University. His current research
interests include all aspects of partial differential equa-
tions and scale-space theory in image analysis. He was
awarded the Wacker Memorial Prize and authored the
book “Anisotropic Diffusion in Image Processing.”
Dr. Joachim Weickert, Department of Computer Science, University of Copen-
hagen, Universitetsparken 1, DK-2100 Copenhagen, Denmark
, />Dieter Willersinn received his diploma in electrical en-
gineering from Technical University Darmstadt in 1988.
From 1988 to 1992 he was with Vitronic Image Process-
ing Systems in Wiesbaden, working on industrial appli-
cations of robot vision and quality control. He then took
a research position at the Technical University in Vienna,
Austria, from which he received his PhD degree in 1995.
In 1995, he joined the Fraunhofer Institute for Informa-
tion and Data Processing (IITB) in Karlsruhe, where he
initially worked on obstacle detection for driver assis-
tance applications. Since 1997, Dr. Willersinn has been
the head of the group, Assessment of Computer Vision
Systems, Department for Recognition and Diagnosis Systems.
Dr. Dieter Willersinn, Fraunhofer Institut IITB, Fraunhoferstr. 1
D-76131 Karlsruhe, Germany,
xxiv Contributors