Content Based Microscopic Image Analysis 1st

Edition Chen Li pdf download

https://ebookname.com/product/content-based-microscopic-image-
analysis-1st-edition-chen-li/

Get Instant Ebook Downloads – Browse at https://ebookname.com

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Partial Differential Equations and Spectral Theory 1st

Edition Li Chen

https://ebookname.com/product/partial-differential-equations-and-
spectral-theory-1st-edition-li-chen/

Geometric Analysis 1st Edition Peter Li

https://ebookname.com/product/geometric-analysis-1st-edition-
peter-li/

Signal and Image Processing for Remote Sensing 1st

Edition C.H. Chen

https://ebookname.com/product/signal-and-image-processing-for-
remote-sensing-1st-edition-c-h-chen/

Orthopaedic and Trauma Nursing An Evidence based

Approach to Musculoskeletal Care 1st Edition Sonya
Clarke

https://ebookname.com/product/orthopaedic-and-trauma-nursing-an-
evidence-based-approach-to-musculoskeletal-care-1st-edition-
sonya-clarke/
CompTIA A Certification All in One Exam Guide Sixth
Edition Michael Meyers

https://ebookname.com/product/comptia-a-certification-all-in-one-
exam-guide-sixth-edition-michael-meyers/

Logic For Dummies Mark Zegarelli

https://ebookname.com/product/logic-for-dummies-mark-zegarelli/

Radar Signal Analysis and Processing Using MATLAB 1st

Edition Bassem R. Mahafza

https://ebookname.com/product/radar-signal-analysis-and-
processing-using-matlab-1st-edition-bassem-r-mahafza/

Agricultural Production Organic Conventional Systems

1st Edition O M Akinyemi

https://ebookname.com/product/agricultural-production-organic-
conventional-systems-1st-edition-o-m-akinyemi/

Maran Illustrated Yoga 1st Edition Marangraphics

Development Group

https://ebookname.com/product/maran-illustrated-yoga-1st-edition-
marangraphics-development-group/
National Electrical Code 1st Edition National Fire
Protection Association

https://ebookname.com/product/national-electrical-code-1st-
edition-national-fire-protection-association/
Studien zur Mustererkennung

Band 39

Chen Li

Content-based Mi ros opi

Image Analysis

λογος
 Content-based Microscopic Image Analysis

DISSERTATION

zur Erlangung des Grades eines Doktors

der Ingenieurwissenschaften

vorgelegt von

M. Sc. Chen Li
geb. am 22.04.1985 in China

eingereicht bei der Naturwissenschaftlich-Technischen Fakultät

der Universität Siegen

Siegen 2015

stand: November 2015

Studien zur Mustererkennung
herausgegeben von:
Prof. Dr.-Ing. Heinrich Niemann
Prof. Dr.-Ing. Elmar Nöth

Bibliografische Information der Deutschen Nationalbibliothek

Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der

Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind
im Internet über http://dnb.d-nb.de abrufbar.

c Copyright Logos Verlag Berlin GmbH 2016

Alle Rechte vorbehalten.

ISBN 978-3-8325-4253-5
ISSN 1617-0695

Logos Verlag Berlin GmbH

Comeniushof
Gubener Str. 47
10243 Berlin
Tel.: +49 030 42 85 10 90
Fax: +49 030 42 85 10 92
INTERNET: http://www.logos-verlag.de
 iii

Gutachter der Dissertation:

1. Prof. Dr.-Ing. Marcin Grzegorzek

2. Prof. Dr.-Ing. Klaus-Dieter Kuhnert

Tag der mündlichen Prüfung:

16. Februar 2016.
iv
 v

Acknowledgements

First of all, I would like to express my utmost gratitude to my first supervisor,

Prof. Dr.-Ing. Marcin Grzegorzek, for his supervision, advice, and guidance on this
research and encouragement and support throughout the time of my study at the
University of Siegen. His broad knowledge in pattern recognition, infinite wisdom,
modesty and spirit inspired and enriched my growth as a student and a researcher
that I want to be, and enlightened me very much to find solutions whenever I
encountered problems during my PhD studies. Then, I wish to give great thanks to
my second supervisor Prof. Dr.-Ing. Klaus-Dieter Kuhnert for his huge help and
support regarding my academic development. And I also thank Prof. Dr. Roland
Wismüller and Prof. Dr.-Ing. Madjid Fathi a lot for their positive evaluation to my
PhD work.

I would like to give my thanks and appreciation to my project leader, Dr. Kimiaki
Shirahama, for his consistent help, invaluable academic guidance and attention
during the entire process of this work. His commitment, scientific knowledge, and
continuous support during the past years were crucial not only for the accomplish-
ment of this work, but also for the expansion of my scientific knowledge and my
growing interest in scientific studies. For me, he is the Mr. Fujino to Mr. Lu Xun.

I want to thank my parents for always encouraging me to pursue whatever interest

I might have fancied at the moment. I am grateful that I have had the opportunity to
find my own path in life and that I have always felt your support. I am also grateful
for the support I have received from my parents-in-law, not only during my PhD
studies. Furthermore, long live all my grandparents and relatives!

The gratitude I feel towards my wife Ning Xu is beyond expression. You are
the best! Without you, I probably would have never even started this journey, and
without you, I would not have finalised it. Thank you for all the joy and support you
have given me. Thank you for the many memorable times staying in Germany with
vi

you.

Thanks to all of my friends, colleagues and coauthors in Germany whom I had a

wonderful time with, and the friends in China who share my failures and successes.
In particular, I thank Prof. Dr. Kuniaki Uehara, Prof. Dr. Oliver Deussen and Prof.
Dr. Beihai Zhou for their great academic support; I thank Prof. Dr. Tao Jiang, Dr.
Minmin Shen and Dr. Fangshu Ma for their excellent long-term cooperation; I thank
Dr.-Ing. Joanna Czajkowska and Dipl.-Inform. Christian Feinen for their friendly
tutoring at the beginning of my PhD study; I thank Dr. Peter Volz, M.Sc. Zeyd
Boukhers, M.Sc. Lukas Köping, Ms. Kristin Klaas, Mr. Florian Schmidt and Mr.
Oliver Tiebe for their helpful technical support in many aspects; I also thank everyone
else, thank you very much for your help of all kinds.

I am thankful for the financial support I received from the China Scholarship
Council, and the graduation aid programme of the DAAD at the University of Siegen.
 vii

Abstract

In this dissertation, novel Content-based Microscopic Image Analysis (CBMIA) methods,

including Weakly Supervised Learning (WSL), are proposed to aid biological studies.
In a CBMIA task, noisy image, image rotation, and object recognition problems need
to be addressed. To this end, the first approach is a general supervised learning
method, which consists of image segmentation, shape feature extraction, classifica-
tion, and feature fusion, leading to a semi-automatic approach. In contrast, the second
approach is a WSL method, which contains Sparse Coding (SC) feature extraction,
classification, and feature fusion, leading to a full-automatic approach. In this WSL
approach, the problems of noisy image and object recognition are jointly resolved
by a region-based classifier, and the image rotation problem is figured out through
SC features. To demonstrate the usefulness and potential of the proposed methods,
experiments are implemented on different practical biological tasks, including envi-
ronmental microorganism classification, stem cell analysis, and insect tracking.

This dissertation is structured as follows: Chapter 1 first prefaces the whole the-
sis with a brief introduction. Then, Chapter 2 reviews related CBMIA techniques
and applications. Next, a semi-automatic microscopic image classification system is
proposed in Chapter 3, using strongly supervised learning methods. In contrast, full-
automatic WSL approaches are introduced in Chapter 4. After the aforementioned
supervised learning methods, an unsupervised learning approach is developed in
Chapter 5 to solve problems regarding microscopic image clustering. Following that,
Chapter 6 introduces a CBMIA framework for the task of multi-object tracking. Then,
the proposed CBMIA methods are tested on three microbiological tasks to prove their
usefulness and effectiveness in Chapter 7, including environmental microorganism
classification, stem cell clustering and insect tracking. Finally, Chapter 8 completes
the whole dissertation with an overall conclusion of the current work and a brief plan
for the future research.
viii

Zusammenfassung
In der vorliegenden Dissertation werden neue Methoden zur inhaltsbasierten,
mikroskopischen Bildanalyse (CBMIA) vorgestellt. Die Ansätze basieren dabei
unter anderem auf Techniken aus dem schwach überwachten Lernen und finden
vor allem Anwendung in biologischen Aufgabenbereichen. In der inhaltsbasierten,
mikroskopischen Bildanalyse müssen Probleme wie verrauschte Bilder, Bildrotation
und Objekterkennung adressiert werden. Hierfür stützt sich der erste vorgestellte
Ansatz auf eine semi-automatisierte Technik, die eine allgemeine, überwachte Lern-
methode darstellt. Der Ansatz beruht auf der Segmentierung von Bildern, auf von der
Objektform abhängigen Merkmalen und auf der Klassifikation und Fusion von Merk-
malen. Im Gegensatz dazu handelt es sich beim zweiten Ansatz um eine schwach
überwachte Lernmethode. Die Merkmalsextraktion erfolgt hierbei durch Sparse Co-
ding (SC). Zusätzlich beinhaltet sie eine Klassifikation und Merkmalsfusion, was
letztlich zu einem voll-automatisierten System führt. Im schwach überwachten Lern-
ansatz werden die zwei Probleme, Bildrauschen und Objekterkennung, gemeinsam
mittels eines auf Bildregionen basierendem Klassifikators gelöst, während das Pro-
blem der Bildrotation mittels SC angegangen wird. Die Nützlichkeit und Effektivität
der vorgestellten Methoden wird anhand verschiedener Aufgaben aus dem Bereich
der Biologie aufgezeigt. Diese Aufgaben beinhalten die Klassifikation von Mikroor-
ganismen, Stammzellenanalyse und die automatisierte Verfolgung von Insekten.
Die Dissertation ist wie folgt aufgebaut: Kapitel 1 führt den Leser in einer
kurzen Einleitung in das Thema ein. In Kapitel 2 folgt eine Aufarbeitung von ver-
wandten Techniken aus dem Bereich der inhaltsbasierten, mikroskopischen Bild-
analyse und seiner Anwendungsbereiche. Anschließend wird in Kapitel 3 ein semi-
automatisiertes System zur Klassifizierung von Mikroskop-Bildern vorgestellt, das
auf Methoden des stark überwachten Lernens basiert. Im Gegensatz dazu wer-
den in Kapitel 4 voll-automatisierte, schwach-überwachte Lernmethoden eingeführt.
Nach den gerade erwähnten überwachten Lernmethoden wird in Kapitel 5 ein
unüberwachter Lernansatz entwickelt, um Probleme bezüglich der mikroskopischen
Bildgruppierung zu lösen. Kapitel 6 führt ein CBMI basiertes Framework zur Multi-
Objekt-Verfolgung ein. Im Anschluss werden in Kapitel 7 die vorgestellten Ansätze
anhand von drei Aufgaben aus der Mikrobiologie auf ihre Effektivität und Nütz-
lichkeit getestet. Die Aufgaben umfassen die Klassifikation von Mikroorganismen,
die Gruppierung von Stammzellen und das automatisierte Verfolgen von Insekten.
Die Dissertation wird in Kapitel 8 durch eine Zusammenfassung der Arbeit und
Aufzeigen zukünftiger Forschungsvorhaben abgeschlossen.
ix
x
CONTENTS xi

Contents

Acknowledgements v

Abstract vii

Contents xi

List of Figures xv

List of Tables xxiii

1 Introduction 1
1.1 Fundamental Concept of Content-based Microscopic Image Analysis . 1
1.2 Motivation of the Present Work . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Contribution of the Present Work . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Overview of this Dissertation . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Related Work 13
2.1 Related Algorithms in CBMIA . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.2 Shape Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.3 Sparse Coding Features . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.4 Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.5 Unsupervised Learning . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.6 Object Detection and Tracking . . . . . . . . . . . . . . . . . . . 22
2.2 Biological Applications Using CBMIA . . . . . . . . . . . . . . . . . . . 26
2.2.1 Microorganism Classification . . . . . . . . . . . . . . . . . . . . 26
2.2.2 Cell Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.3 Insect Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
xii CONTENTS

3 Semi-automatic Microscopic Image Classification Using Strongly Super-

vised Learning 37
3.1 Overview of the System . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.1 Sobel Edge Detection and Morphological Operations . . . . . . 39
3.2.2 Semi-automatic Image Segmentation . . . . . . . . . . . . . . . 45
3.3 Global Shape Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.1 Isomerous Edge Histogram Descriptor . . . . . . . . . . . . . . 47
3.3.2 Basic Geometrical Features . . . . . . . . . . . . . . . . . . . . . 48
3.3.3 Fourier Descriptor . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3.4 Internal Structure Histogram . . . . . . . . . . . . . . . . . . . . 58
3.4 Local Shape Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5 Strongly Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5.1 Fundamentals of SVM . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5.2 Radial Basis Function Kernel SVM . . . . . . . . . . . . . . . . . 62
3.5.3 Late Fusion of SVM . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4 Full-automatic Microscopic Image Classification Using Weakly Supervised

Learning 65
4.1 Overview of the System . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Sparse Coding Features . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 Sparse Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.2.2 Non-negative Sparse Coding . . . . . . . . . . . . . . . . . . . . 70
4.3 Weakly Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3.1 Basic RBSVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.2 Improved RBSVM . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

5 Microscopic Image Clustering Using Unsupervised Learning 77

5.1 Overview of the System . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Full-automatic Image Segmentation . . . . . . . . . . . . . . . . . . . . 79
5.2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.2.2 Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 Global Shape Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.3.1 Edge Histogram Descriptor . . . . . . . . . . . . . . . . . . . . . 81
5.3.2 Higher-level Geometrical Feature . . . . . . . . . . . . . . . . . 83
5.3.3 Shape Signature Histogram . . . . . . . . . . . . . . . . . . . . . 85
5.3.4 Shape Context Feature . . . . . . . . . . . . . . . . . . . . . . . . 85
CONTENTS xiii

5.4 Unsupervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

5.4.1 k-means Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4.2 Clustering Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 87
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6 Object Tracking Using Interactive Learning 91

6.1 Fundamental Conception . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2.1 Object Detection and Preprocessing . . . . . . . . . . . . . . . . 96
6.2.2 Anatomical Model of Insect Body Parts . . . . . . . . . . . . . . 96
6.3 Interactive Object Tracking Framework . . . . . . . . . . . . . . . . . . 98
6.3.1 Object Classification . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.3.2 Constrained Frame-to-Frame Linking . . . . . . . . . . . . . . . 100
6.3.3 Interactive KF Estimation and Annotation Query . . . . . . . . 101
6.3.4 Track Linking Through Merge Conditions . . . . . . . . . . . . . 103
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

7 Applications and Experiments 107

7.1 Environmental Microorganism Classification . . . . . . . . . . . . . . . 107
7.1.1 EM Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.1.2 Classification Using SSL . . . . . . . . . . . . . . . . . . . . . . . 110
7.1.3 Classification Using WSL . . . . . . . . . . . . . . . . . . . . . . 117
7.2 Stem Cell Analysis and Clustering . . . . . . . . . . . . . . . . . . . . . 120
7.2.1 Evaluation of Segmentation . . . . . . . . . . . . . . . . . . . . . 121
7.2.2 Evaluation of Clustering . . . . . . . . . . . . . . . . . . . . . . . 122
7.3 Insect Body Parts Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.3.1 Dataset and Experimental Setting . . . . . . . . . . . . . . . . . 125
7.3.2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.4 Extended Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
7.4.1 CBMIR on EMDS Images . . . . . . . . . . . . . . . . . . . . . . 132
7.4.2 Late Fusion of Global Shape Features on MPEG-7 Dataset . . . 134
7.4.3 EM Classification Using Pair-wise Local Features . . . . . . . . 137
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

8 Conclusion and Future Work 141

8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

List of Publications 169

xiv CONTENTS
LIST OF FIGURES xv

List of Figures

1.1 Examples of CBMIA domains. (a) Micro-operation [Son+06]: Two joy

sticks are used to handle a micro-component. (b) Soil structure analysis:
A piece of soil shows a typical granular structure under a microscope.
(c) Plant tissue analysis: Structure of cell wall of a dry poplar leaf is
observed. (d) Rock analysis [LM14]: The microscopic structure of a
rock is revealed, where different elements show different colours. (e)
Cell tissue analysis: Stem cells that are in a migration are recorded by
microscopic images. (f) Microorganism analysis: Epistylis in activated
sludge is checked to evaluate the pollution level of the environment. . 2
1.2 Examples of problems addressed by CBMIA. (a) An undirected image
of Dicranophorus with a lot of noise and impurities. (b) A stem cell with
very thin cytoplasm that is easy to over-segment. (c) The same plant
leaf under natural (left) and artificial (right) light sources, which shows
totally different colours in pink and orange. (d) An image of transparent
Noctiluca which has no obvious texture and colour characteristics. (e)
An image of detected Stentor (in the red bounding box), excluding
impurities at the bottom right corner. (f) Two fast moving antennas (in
the red and blue bounding boxes) of a honey bee are tracked. . . . . . 4
1.3 The general working flow of a CBMIA system. . . . . . . . . . . . . . . 7
1.4 Examples of EMs in their corresponding environments. (a) Epistylis in
sludge. (b) Rotifera in a river. (c) Vorticella in a wastewater plant. . . . 8
1.5 An example of different structures of objects. (a) A rounded structure
of Arcella. (b) A radiate structure of Actinophrys. . . . . . . . . . . . . . 9

3.1 An overview of the EM classification system using SSL framework. . . 38

xvi LIST OF FIGURES

3.2 An example of gradient theory for edge detection. (a) A brighter edge
between two darker parts. (b) A darker edge between two brighter
parts. The top row shows the original images. The middle row shows
the overall changing trends of the pixel values ignoring noise, where
brighter parts have higher pixel values than darker parts. The bottom
row shows the locations where the edges occur. . . . . . . . . . . . . . 41
3.3 Mask matrixes of Sobel edge detectors. (a) The mask matrix for x-
direction. (b) The mask matrix for y-direction. . . . . . . . . . . . . . . 41
3.4 An example for pixel values in an image patch. The left image shows
an original image of Vorticella, where nine pixels in red boxes constitute
an image patch. The left matrix shows the pixel values of the left image
patch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 An example for the usage of mask matrix in Soble edge detection.
The left image shows the working region of the masks in an image of
Vorticella. The right image shows the objective pixel of the masks. . . . 42
3.6 An example of Sobel edge detection. (a) An original image of Vorticella.
(b) Edge detection result using the Sobel method. . . . . . . . . . . . . 43
3.7 An example of a median filter for image denoising. The top left matrix
shows the pixel values of an original image patch. The bottom array is
the ranked pixel values. The top right matrix shows the pixel values
of the filtered image patch. In the original patch, the central pixel (in
the blue box) is a noisy pixel whose value is obviously higher than its
surrounding pixels. By the filtering operation, the value of this pixel is
replaced by the median value of all the pixels (in the red box.) . . . . . 44
3.8 An example of an opening operation. (a) An original image. (b) The
image after erosion. (c) The image after dilation. . . . . . . . . . . . . . 44
3.9 An example of a closing operation. (a) An original image. (b) The
image after dilation. (c) The image after erosion. . . . . . . . . . . . . . 45
3.10 An example for a filling holes operation. (a) An original image with a
blank hole in an object. (b) The image after filling holes operation. . . 45
3.11 An example for the proposed semi-automatic segmentation approach.
(a) An original microscopic image of Rotifera. (b) Manual region se-
lection. (c) Sobel edge detection. (d) Morphological operations. (e)
Manual object selection. (f) The final segmentation result. . . . . . . . 46
3.12 An example of the continuous degree of edges. The left image is the
boundary of Paramecium. The right image shows an amplified part of
the boundary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
LIST OF FIGURES xvii

3.13 An example of IEHD. (a) An image of Actinophrys. (b) An image of

Arcella. (c) IEHDs of (a) and (b). The blue bins represent the IEHD of
(a), and the red bins show the IEHD of (b). The horizontal axis shows
the 13 intervals of edges. The vertical axis shows the amount of edges. 48
3.14 An example of the complexity. From left to right, four EM images show
Actinophrys, Aspidisca, Paramecium and Arcella, respectively. The value
under each image is the complexity of the above mentioned EM. . . . 49
3.15 An example of Hotelling transformation for image rotation. (a) is the
original image of Rotifera. (b), (c), and (d) are the rotated results of
(a) with 90◦ , 180◦ , and 270◦ , separately. (e), (f), (g), and (h) are the
rotation results of the Hotelling transformation of (a), (b), (c), and (d),
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.16 An example of the discriminative power of SS. (a) and (c) are the images
of Actinophrys and Euglypha, respectively. (b) and (d) are the SSs of (a)
and (c), separately. The horizontal axis in (b) and (d) shows the points
on the contour of each object, and the vertical axis shows the centriod
distances of these two objects, respectively. . . . . . . . . . . . . . . . . 56
3.17 An example of the rotation problem of SS. (a) and (c) are the original
and rotated (180◦ ) images of Rotifera, respectively. (b) and (d) are the
SSs of (a) and (c), separately. The setting of horizontal and vertical axis
in (b) and (d) is the same as that in Figure 3.16. . . . . . . . . . . . . . . 57
3.18 An illustration of discriminative properties of ISH. The original image
of Rotifera in (a) is scaled, rotated and deformed in (b), (c) and (d),
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.19 An example of SVMs. (a) is a linear SVM. (b) shows an RBFSVM. . . . 61

4.1 An overview of the EM classification system using WSL framework. . 67

4.2 An example of image reconstruction by the linear combination of bases. 68
4.3 An example of the over-complete set of basis elements. . . . . . . . . . 68
4.4 An example of SC for image reconstruction and denoising. (a) is an
original image of Rotifera. (b) shows the reconstruction result of (a). (c)
is the denoising result of (a). . . . . . . . . . . . . . . . . . . . . . . . . 70
4.5 An example of the basic RBSVM using BoVW features. . . . . . . . . . 72
4.6 An example of the improved RBSVM using NNSC features. . . . . . . 74
xviii LIST OF FIGURES

5.1 Working flow of the proposed stem cell clustering system. (a) is an
original stem cell microscopic image. (b) shows the segmented stem
cells. (c) is the feature extraction approach. (d) shows feature selection
and fusion. (e) is the final classification result, where cells with similar
shapes are classified into the same category. . . . . . . . . . . . . . . . . 78
5.2 An example of the double-stage segmentation approach. (a) shows an
original microscopic image of stem cells. (b) is the segmentation result
of the first stage. (c) shows the segmentation result of the second stage.
(d) is the final segmentation result combined by (b) and (c). . . . . . . 80
5.3 Five types of edges. (a) Vertical (0◦ ) edge. (b) 45◦ edges. (c) Horizontal
(90◦ ) edge. (d) 135◦ edge. (e) Uncertain edge. . . . . . . . . . . . . . . 81
5.4 An example of EHD. (a) An image of Actinophrys. (b) An image of
Aspidisca. (c) EHDs of (a) and (b). The blue bins represent the EHD of
(a), and the red bins show the EHD of (b). The horizontal axis shows
the five types of edges. The vertical axis shows the amount of edges. . 82
5.5 An example of EHD in different directions. (a) An image of Paramecium
(deviation to the right). (b) An image of Paramecium (deviation to the
left). (c) EHDs of (a) and (b), where the horizontal axis shows the five
types of edges. (d) Modified EHDs of (a) and (b), where the horizontal
axis shows edges in an ascending order. The blue bins represent the
edges of (a), and the red bins show the edges of (b). The vertical axis
displays the amount of edges. . . . . . . . . . . . . . . . . . . . . . . . 82
5.6 An example of Hu moments. (a) is the original image of Ceratium. (b),
(c), and (d) are the deformed results of (a) with translation, 0.5 time
scaled and 180◦ rotation, respectively. (e), (f), (g), and (h) are the visible
histograms of Hu moments of (a), (b), (c), and (d), separately. The
horizontal axis in each histogram shows seven Hu moments, and the
vertical axis shows their normalised values in [0, 1]. . . . . . . . . . . . 84
5.7 An example of SCF on an image of Synchaeta. . . . . . . . . . . . . . . 86
5.8 An example of k-means clustering. The red, orange and blue points
show three classes of data. The purple points are the clustering centres
of these three classes, respectively. The red, orange and blue circles
represent the ranges of Euclidean distances of these three classes. . . . 87
5.9 An example of the silhouette plot for clustering evaluation. The hori-
zontal axis shows the value of the silhouette plot’s measurement. The
vertical axis shows the clustering numbers of points in each cluster. . 88
LIST OF FIGURES xix

6.1 Sample frames of (a, e) merge or (b, c, d) occlusion. Merged targets are
difficult to be differentiated at bounding box level, thus it is proposed
to estimate the position of the tip of each target, which is denoted as a
solid circle in the corresponding colour. . . . . . . . . . . . . . . . . . . 92
6.2 Object detections at ten consecutive frames including merged and false
negative bounding boxes. The identification of each bounding box,
shown in a different colour, is a challenging task. The label for each
body part is denoted as 1: Left antenna; 2: Left mandible; 3: Proboscis;
4: Right mandible; 5: Right antenna. . . . . . . . . . . . . . . . . . . . . 92
6.3 The flowchart of the overall tracking framework: the yellow blocks
highlight the interactive part, while the blue blocks denote the auto-
mated computation part. . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.4 Associative odor-reward learning paradigm in honey bees. A bee that
has learnt the association between an odorant and a food-reward ex-
tends its proboscis when stimulated with the learnt odorant: (a) Before
odorant stimulation. (b) Odorant released indicated by the LED. (c)
Sugar rewarding. (d) During odorant stimulation. . . . . . . . . . . . . 95
6.5 The closed up sample video frame of an ant. In the case of an ant,
two antennae (yellow and blue bounding boxes) and its mouthparts
(purple and green bounding boxes) need to be tracked (i.e. n = 4). . . . 97
6.6 The initial interest points in (a) (Denoted by blue stars) are detected
by a corner detector within the green bounding box. The number of
successfully tracked points reduces dramatically over time (b) ten, (c)
six and (d) zero. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.7 The first row are detected antennae, the second row are detected
mandibles, and the third row are detected proboscis. . . . . . . . . . . 98
6.8 An example of the voting process of 1vs1 MSVM, where ω2 obtains the
highest voting number of 4 and is identified as the final classification
result. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.9 An example of linking tracks through merge condition: the shaded
lines indicate the tracks at bounding box level, and the circles indicate
p p
the tips. t31 , t32 indicate the tail and head of the pth tracklet of the
proboscis (i.e. label 3) T3p p , respectively. . . . . . . . . . . . . . . . . . 103
t31 ,t32

7.1 An example of EM images in EMDS (EMDS-4). (a) shows the original

images. (b) shows the ground truth images. . . . . . . . . . . . . . . . 109
7.2 EM classification results on original and ground truth images, where
‘O’ and ‘G’ on the right side of feature names represent that features
are extracted from original and ground truth images, respectively. . . . 111
xx LIST OF FIGURES

7.3 EM classification results on BGF and ISH, where ‘F’, ‘GC’ and ‘S’ on the
right side of feature names represent that features are extracted from
segmentation results by the full-automatic, GrabCut and the proposed
semi-automatic methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.4 Examples of segmentation results. From left to right, each row shows
the original and ground truth images of EMs (Aspidisca, Codosiga and
Ceratium), the segmentation results by the full-automatic method, the
results by GrabCut, and the results by the proposed semi-automatic
method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.5 EM classification results of five features on ground truth images. . . . 115
7.6 EM classification results of five features on results by the proposed
semi-automatic segmentation method. . . . . . . . . . . . . . . . . . . . 115
7.7 EM classification results using early fusion, late fusion and MKL. ‘G’
means using ground truth images, and ‘S’ means using the segmenta-
tion results by the proposed semi-automatic method. . . . . . . . . . . 117
7.8 The localisation results of the improved RBSVM. . . . . . . . . . . . . 120
7.9 EM classification results using late fusion in the WSL framework. . . . 121
7.10 A comparison of the double-stage segmentation to other methods. (a)
An original image; (b) The ground truth image; (c) Double-stage; (d)
Watershed; (e) Region Growing; (f) Thresholding; (g) Normalised Cut;
(h) Contour detection and hierarchical image segmentation. Different
colours are used to distinguish neighbouring segmented cells. . . . . . 123
7.11 Silhouette plot of stem cell clustering. . . . . . . . . . . . . . . . . . . . 125
7.12 Two snapshots of the GUI: Initial tracking hypothesis on a KF (a) and
user corrected labels (b). LocoTracker enables users to correct tracking
errors including mismatches, false positives and false negatives. . . . 126
7.13 (a) KF ratio vs. Iteration number of ten tested videos: The user query
stops at a KF ratio of 0.1 ∼ 0.18, and the KF ratio drops dramatically
within five iterations. (b) Tracking error vs. annotation time: The
TE for all bee videos drops below 0.05 at the final interaction, while
additional annotation time is about one second on each frame. . . . . 128
7.14 Results of four tracking methods. . . . . . . . . . . . . . . . . . . . . . . 129
7.15 Ten consecutive sample frames of the final tracking results under merge
condition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.16 Three trajectories of tips of 100 frames in the videos shown in Figure 6.4
are drawn on one of the video frames: The orange dots denote the tips
of the right antenna, red for the proboscis and blue for the left antenna. 131
7.17 Working flow of the proposed EM-CBMIR system. . . . . . . . . . . . . 134
LIST OF FIGURES xxi

7.18 Comparison of EM-CBMIR results using ISH on ground truth (GT) and
semi-automatic segmented (S-Seg) images. The horizontal axis shows
the AP and mean AP (µ). The vertical axis shows the value of AP and
mean AP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.19 Examples of EM-CBMIR results using ISH. The first column shows the
query images. From the second to the last column, the database images
are sorted by their ISH similarities from high to low. The images in red
bounding boxes are the relevant images. . . . . . . . . . . . . . . . . . 135
7.20 An example of the MPEG-7 dataset. . . . . . . . . . . . . . . . . . . . . 136
7.21 EM classification results of PSIFT and MPSIFT features. SIFT & PSIFT
shows the classification result of early fusion of SIFT and PSIFT. Simi-
larly, SIFT & MPSIFT shows that of SIFT and MPSIFT. . . . . . . . . . 139
xxii LIST OF FIGURES
LIST OF TABLES xxiii

List of Tables

2.1 Object tracking related work and main characteristics. . . . . . . . . . . 25

2.2 A comparison of traditional methods for microscopic image analysis. . 26

7.1 Information of EM categories. . . . . . . . . . . . . . . . . . . . . . . . . 108

7.2 Evaluation of segmentation results in terms of similarity Si, sensitivity
Se, and specificity Sp (given in [%]). ‘F’, ‘G’ and ‘S’ in parenthesis rep-
resent the full-automatic, GrabCut and the proposed semi-automatic
methods, respectively. µ indicates the mean value over 15 classes. . . . 113
7.3 Fusion weights obtained by the proposed late fusion method on semi-
segmented EM images. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.4 EM Classification results using NNSC and BoVW features by lin-
ear SVM and RBSVM. APs and the mean of APs (µ) are in [%].
‘(S+N)’ shows (SVM+NNSC), ‘(S+B)’ means (SVM+BoVW), ‘(R+N)’
is (RBSVM+NNSC) and ‘(R+B)’ represents (RBSVM+BoVW). . . . . . . 119
7.5 A comparison of the double-stage segmentation to other methods in
terms of similarity Si, sensitivity Se, specificity Sp and ratio Ra of seg-
mented cells and existing cells (given in [%]). . . . . . . . . . . . . . . 122
7.6 Mean variance evaluation of stem cell clustering results. The first
column shows different global shape features. The first row shows
different k numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.7 The characteristics of tested videos. . . . . . . . . . . . . . . . . . . . . . 126
7.8 TE of three methods on different insect videos. . . . . . . . . . . . . . . 128
7.9 Comparison of proposed method with ground truth. . . . . . . . . . . 130
7.10 Classification results in Section 6.3.1 for six global shape features. . . . 132
7.11 Retrieval results on MPEG-7 dataset using the proposed global shape
features and PSSGM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
 Another Random Document on
Scribd Without Any Related Topics
that usually terminated the exploits of Arsène Lupin. She remained
standing near the door, surprised at the silence, and looked about
her without any display of suspicion or fear.
“She will get away! She will disappear!” thought Ganimard.
Then he managed to get between her and the door. She turned to
go out.
“No, no!” he said. “Why are you going away?”
“Really, monsieur, I do not understand what this means. Allow me
—”
“There is no reason why you should go, madame, and very good
reasons why you should remain.”
“But—”
“It is useless, madame. You cannot go.”
Trembling, she sat on a chair, and stammered:
“What is it you want?”
Ganimard had won the battle and captured the blonde Lady. He
said to her:
“Allow me to present the friend I mentioned, who desires to
purchase some diamonds. Have you procured the stones you
promised to bring?”
“No—no—I don’t know. I don’t remember.”
“Come! Jog your memory! A person of your acquaintance intended
to send you a tinted stone.... ‘Something like the blue diamond,’ I
said, laughing; and you replied: ‘Exactly, I expect to have just what
you want.’ Do you remember?”
She made no reply. A small satchel fell from her hand. She picked
it up quickly, and held it securely. Her hands trembled slightly.
“Come!” said Ganimard, “I see you have no confidence in us,
Madame de Réal. I shall set you a good example by showing you
what I have.”
He took from his pocketbook a paper which he unfolded, and
disclosed a lock of hair.
 “These are a few hairs torn from the head of Antoinette Bréhat by
the Baron d’Hautrec, which I found clasped in his dead hand. I have
shown them to Mlle. Gerbois, who declares they are of the exact
color of the hair of the blonde Lady. Besides, they are exactly the
color of your hair—the identical color.”
Madame Réal looked at him in bewilderment, as if she did not
understand his meaning. He continued:
“And here are two perfume bottles, without labels, it is true, and
empty, but still sufficiently impregnated with their odor to enable
Mlle. Gerbois to recognize in them the perfume used by that blonde
Lady who was her traveling companion for two weeks. Now, one of
these bottles was found in the room that Madame de Réal occupied
at the Château de Crozon, and the other in the room that you
occupied at the Hôtel Beaurivage.”
“What do you say?... The blonde Lady ... the Château de
Crozon....”
The detective did not reply. He took from his pocket and placed on
the table, side by side, four small sheets of paper. Then he said:
“I have, on these four pieces of paper, various specimens of
handwriting; the first is the writing of Antoinette Bréhat; the second
was written by the woman who sent the note to Baron Herschmann
at the auction sale of the blue diamond; the third is that of Madame
de Réal, written while she was stopping at the Château de Crozon;
and the fourth is your handwriting, madame ... it is your name and
address, which you gave to the porter of the Hôtel Beaurivage at
Trouville. Now, compare the four handwritings. They are identical.”
“What absurdity is this? really, monsieur, I do not understand.
What does it mean?”
“It means, madame,” exclaimed Ganimard, “that the blonde Lady,
the friend and accomplice of Arsène Lupin, is none other than you,
Madame Réal.”
Ganimard went to the adjoining room and returned with Mon.
Gerbois, whom he placed in front of Madame Réal, as he said:
 “Monsieur Gerbois, is this the person who abducted your daughter,
the woman you saw at the house of Monsieur Detinan?”
“No.”
Ganimard was so surprised that he could not speak for a moment;
finally, he said: “No?... You must be mistaken....”
“I am not mistaken. Madame is blonde, it is true, and in that
respect resembles the blonde Lady; but, in all other respects, she is
totally different.”
“I can’t believe it. You must be mistaken.”
Ganimard called in his other witnesses.
“Monsieur d’Hautrec,” he said, “do you recognize Antoinette
Bréhat?”
“No, this is not the person I saw at my uncle’s house.”
“This woman is not Madame de Réal,” declared the Count de
Crozon.
That was the finishing touch. Ganimard was crushed. He was
buried beneath the ruins of the structure he had erected with so
much care and assurance. His pride was humbled, his spirit was
broken, by the force of this unexpected blow.
Mon. Dudouis arose, and said:
“We owe you an apology, madame, for this unfortunate mistake.
But, since your arrival here, I have noticed your nervous agitation.
Something troubles you; may I ask what it is?”
“Mon Dieu, monsieur, I was afraid. My satchel contains diamonds
to the value of a hundred thousand francs, and the conduct of your
friend was rather suspicious.”
“But you were frequently absent from Paris. How do you explain
that?”
“I make frequent journeys to other cities in the course of my
business. That is all.”
Mon. Dudouis had nothing more to ask. He turned to his
subordinate, and said:
 “Your investigation has been very superficial, Ganimard, and your
conduct toward this lady is really deplorable. You will come to my
office to-morrow and explain it.”
The interview was at an end, and Mon. Dudouis was about to
leave the room when a most annoying incident occurred. Madame
Réal turned to Ganimard, and said:
“I understand that you are Monsieur Ganimard. Am I right?”
“Yes.”
“Then, this letter must be for you. I received it this morning. It
was addressed to ‘Mon. Justin Ganimard, care of Madame Réal.’ I
thought it was a joke, because I did not know you under that name,
but it appears that your unknown correspondent knew of our
rendezvous.”
Ganimard was inclined to put the letter in his pocket unread, but
he dared not do so in the presence of his superior, so he opened the
envelope and read the letter aloud, in an almost inaudible tone:

“Once upon a time, there were a blonde Lady, a Lupin,

and a Ganimard. Now, the wicked Ganimard had evil
designs on the pretty blonde Lady, and the good Lupin
was her friend and protector. When the good Lupin
wished the blonde Lady to become the friend of the
Countess de Crozon, he caused her to assume the
name of Madame de Réal, which is a close
resemblance to the name of a certain diamond broker,
a woman with a pale complexion and golden hair. And
the good Lupin said to himself: If ever the wicked
Ganimard gets upon the track of the blonde Lady, how
useful it will be to me if he should be diverted to the
track of the honest diamond broker. A wise precaution
that has borne good fruit. A little note sent to the
newspaper read by the wicked Ganimard, a perfume
bottle intentionally forgotten by the genuine blonde
Lady at the Hôtel Beaurivage, the name and address of
Madame Réal written on the hotel register by the
 genuine blonde Lady, and the trick is played. What do
you think of it, Ganimard? I wished to tell you the true
story of this affair, knowing that you would be the first
to laugh over it. Really, it is quite amusing, and I have
enjoyed it very much.
“Accept my best wishes, dear friend, and give my
kind regards to the worthy Mon. Dudouis.

“ARSÈNE LUPIN.”

“He knows everything,” muttered Ganimard, but he did not see

the humor of the situation as Lupin had predicted. “He knows some
things I have never mentioned to any one. How could he find out
that I was going to invite you here, chief? How could he know that I
had found the first perfume bottle? How could he find out those
things?”
He stamped his feet and tore his hair—a prey to the most tragic
despair. Mon. Dudouis felt sorry for him, and said:
“Come, Ganimard, never mind; try to do better next time.”
And Mon. Dudouis left the room, accompanied by Madame Réal.

During the next ten minutes, Ganimard read and re-read the letter
of Arsène Lupin. Monsieur and Madame de Crozon, Monsieur
d’Hautrec and Monsieur Gerbois were holding an animated
discussion in a corner of the room. At last, the count approached the
detective, and said:
“My dear monsieur, after your investigation, we are no nearer the
truth than we were before.”
“Pardon me, but my investigation has established these facts: that
the blonde Lady is the mysterious heroine of these exploits, and that
Arsène Lupin directed them.”
 “Those facts do not solve the mystery; in fact, they render it more
obscure. The blonde Lady commits a murder in order to steal the
blue diamond, and yet she does not steal it. Afterward she steals it
and gets rid of it by secretly giving it to another person. How do you
explain her strange conduct?”
“I cannot explain it.”
“Of course; but, perhaps, someone else can.”
“Who?”
The Count hesitated, so the Countess replied, frankly:
“There is only one man besides yourself who is competent to
enter the arena with Arsène Lupin and overcome him. Have you any
objection to our engaging the services of Herlock Sholmes in this
case?”
Ganimard was vexed at the question, but stammered a reply:
“No ... but ... I do not understand what——”
“Let me explain. All this mystery annoys me. I wish to have it
cleared up. Monsieur Gerbois and Monsieur d’Hautrec have the same
desire, and we have agreed to send for the celebrated English
detective.”
“You are right, madame,” replied the detective, with a loyalty that
did him credit, “you are right. Old Ganimard is not able to overcome
Arsène Lupin. But will Herlock Sholmes succeed? I hope so, as I
have the greatest admiration for him. But ... it is improbable.”
“Do you mean to say that he will not succeed?”
“That is my opinion. I can foresee the result of a duel between
Herlock Sholmes and Arsène Lupin. The Englishman will be
defeated.”
“But, in any event, can we count on your assistance?”
“Quite so, madame. I shall be pleased to render Monsieur Sholmes
all possible assistance.”
“Do you know his address?”
“Yes; 219 Parker street.”
 That evening Monsieur and Madame de Crozon withdrew the
charge they had made against Herr Bleichen, and a joint letter was
addressed to Herlock Sholmes.
 CHAPTER III.
HERLOCK SHOLMES OPENS
HOSTILITIES.

“What does monsieur wish?”

“Anything,” replied Arsène Lupin, like a man who never worries
over the details of a meal; “anything you like, but no meat or
alcohol.”
The waiter walked away, disdainfully.
“What! still a vegetarian?” I exclaimed.
“More so than ever,” replied Lupin.
“Through taste, faith, or habit?”
“Hygiene.”
“And do you never fall from grace?”
“Oh! yes ... when I am dining out ... and wish to avoid being
considered eccentric.”
We were dining near the Northern Railway station, in a little
restaurant to which Arsène Lupin had invited me. Frequently he
would send me a telegram asking me to meet him in some obscure
restaurant, where we could enjoy a quiet dinner, well served, and
which was always made interesting to me by his recital of some
startling adventure theretofore unknown to me.
On that particular evening he appeared to be in a more lively
mood than usual. He laughed and joked with careless animation,
and with that delicate sarcasm that was habitual with him—a light
and spontaneous sarcasm that was quite free from any tinge of
malice. It was a pleasure to find him in that jovial mood, and I could
not resist the desire to tell him so.
“Ah! yes,” he exclaimed, “there are days in which I find life as
bright and gay as a spring morning; then life seems to be an infinite
treasure which I can never exhaust. And yet God knows I lead a
careless existence!”
“Too much so, perhaps.”
“Ah! but I tell you, the treasure is infinite. I can spend it with a
lavish hand. I can cast my youth and strength to the four winds of
Heaven, and it is replaced by a still younger and greater force.
Besides, my life is so pleasant!... If I wished to do so, I might
become—what shall I say?... An orator, a manufacturer, a
politician.... But, I assure you, I shall never have such a desire.
Arsène Lupin, I am; Arsène Lupin, I shall remain. I have made a vain
search in history to find a career comparable to mine; a life better
filled or more intense.... Napoleon? Yes, perhaps.... But Napoleon,
toward the close of his career, when all Europe was trying to crush
him, asked himself on the eve of each battle if it would not be his
last.”
Was he serious? Or was he joking? He became more animated as
he proceeded:
“That is everything, do you understand, the danger! The
continuous feeling of danger! To breathe it as you breathe the air, to
scent it in every breath of wind, to detect it in every unusual
sound.... And, in the midst of the tempest, to remain calm ... and
not to stumble! Otherwise, you are lost. There is only one sensation
equal to it: that of the chauffeur in an automobile race. But that race
lasts only a few hours; my race continues until death!”
“What fantasy!” I exclaimed. “And you wish me to believe that you
have no particular motive for your adoption of that exciting life?”
“Come,” he said, with a smile, “you are a clever psychologist. Work
it out for yourself.”
He poured himself a glass of water, drank it, and said:
 “Did you read ‘Le Temps’ to-day?”
“No.”
“Herlock Sholmes crossed the Channel this afternoon, and arrived
in Paris about six o’clock.”
“The deuce! What is he coming for?”
“A little journey he has undertaken at the request of the Count
and Countess of Crozon, Monsieur Gerbois, and the nephew of
Baron d’Hautrec. They met him at the Northern Railway station, took
him to meet Ganimard, and, at this moment, the six of them are
holding a consultation.”
Despite a strong temptation to do so, I had never ventured to
question Arsène Lupin concerning any action of his private life,
unless he had first mentioned the subject to me. Up to that moment
his name had not been mentioned, at least officially, in connection
with the blue diamond. Consequently, I consumed my curiosity in
patience. He continued:
“There is also in ‘Le Temps’ an interview with my old friend
Ganimard, according to whom a certain blonde lady, who should be
my friend, must have murdered the Baron d’Hautrec and tried to rob
Madame de Crozon of her famous ring. And—what do you think?—
he accuses me of being the instigator of those crimes.”
I could not suppress a slight shudder. Was this true? Must I
believe that his career of theft, his mode of existence, the logical
result of such a life, had drawn that man into more serious crimes,
including murder? I looked at him. He was so calm, and his eyes had
such a frank expression! I observed his hands: they had been
formed from a model of exceeding delicacy, long and slender;
inoffensive, truly; and the hands of an artist....
“Ganimard has pipe-dreams,” I said.
“No, no!” protested Lupin. “Ganimard has some cleverness; and,
at times, almost inspiration.”
“Inspiration!”
 “Yes. For instance, that interview is a master-stroke. In the first
place, he announces the coming of his English rival in order to put
me on my guard, and make his task more difficult. In the second
place, he indicates the exact point to which he has conducted the
affair in order that Sholmes will not get credit for the work already
done by Ganimard. That is good warfare.”
“Whatever it may be, you have two adversaries to deal with, and
such adversaries!”
“Oh! one of them doesn’t count.”
“And the other?”
“Sholmes? Oh! I confess he is a worthy foe; and that explains my
present good humor. In the first place, it is a question of self-
esteem; I am pleased to know that they consider me a subject
worthy the attention of the celebrated English detective. In the next
place, just imagine the pleasure a man, such as I, must experience
in the thought of a duel with Herlock Sholmes. But I shall be obliged
to strain every muscle; he is a clever fellow, and will contest every
inch of the ground.”
“Then you consider him a strong opponent?”
“I do. As a detective, I believe, he has never had an equal. But I
have one advantage over him; he is making the attack and I am
simply defending myself. My rôle is the easier one. Besides, I am
familiar with his method of warfare, and he does not know mine. I
am prepared to show him a few new tricks that will give him
something to think about.”
He tapped the table with his fingers as he uttered the following
sentences, with an air of keen delight:
“Arsène Lupin against Herlock Sholmes.... France against
England.... Trafalgar will be revenged at last.... Ah! the rascal ... he
doesn’t suspect that I am prepared ... and a Lupin warned—”
He stopped suddenly, seized with a fit of coughing, and hid his
face in his napkin, as if something had stuck in his throat.
“A bit of bread?” I inquired. “Drink some water.”
 “No, it isn’t that,” he replied, in a stifled voice.
“Then, what is it?”
“The want of air.”
“Do you wish a window opened?”
“No, I shall go out. Give me my hat and overcoat, quick! I must
go.”
“What’s the matter?”
“The two gentlemen who came in just now.... Look at the taller
one ... now, when we go out, keep to my left, so he will not see
me.”
“The one who is sitting behind you?”
“Yes. I will explain it to you, outside.”
“Who is it?”
“Herlock Sholmes.”
He made a desperate effort to control himself, as if he were
ashamed of his emotion, replaced his napkin, drank a glass of water,
and, quite recovered, said to me, smiling:
“It is strange, hein, that I should be affected so easily, but that
unexpected sight—”
“What have you to fear, since no one can recognize you, on
account of your many transformations? Every time I see you it
seems to me your face is changed; it’s not at all familiar. I don’t
know why.”
“But he would recognize me,” said Lupin. “He has seen me only
once; but, at that time, he made a mental photograph of me—not of
my external appearance but of my very soul—not what I appear to
be but just what I am. Do you understand? And then ... and then....
I did not expect to meet him here.... Such a strange encounter!... in
this little restaurant....”
“Well, shall we go out?”
“No, not now,” said Lupin.
“What are you going to do?”
 “The better way is to act frankly ... to have confidence in him—
trust him....”
“You will not speak to him?”
“Why not? It will be to my advantage to do so, and find out what
he knows, and, perhaps, what he thinks. At present I have the
feeling that his gaze is on my neck and shoulders, and that he is
trying to remember where he has seen them before.”
He reflected a moment. I observed a malicious smile at the corner
of his mouth; then, obedient, I think, to a whim of his impulsive
nature, and not to the necessities of the situation, he arose, turned
around, and, with a bow and a joyous air, he said:
“By what lucky chance? Ah! I am delighted to see you. Permit me
to introduce a friend of mine.”
For a moment the Englishman was disconcerted; then he made a
movement as if he would seize Arsène Lupin. The latter shook his
head, and said:
“That would not be fair; besides, the movement would be an
awkward one and ... quite useless.”
The Englishman looked about him, as if in search of assistance.
“No use,” said Lupin. “Besides, are you quite sure you can place
your hand on me? Come, now, show me that you are a real
Englishman and, therefore, a good sport.”
This advice seemed to commend itself to the detective, for he
partially rose and said, very formally:
“Monsieur Wilson, my friend and assistant—Monsieur Arsène
Lupin.”
Wilson’s amazement evoked a laugh. With bulging eyes and
gaping mouth, he looked from one to the other, as if unable to
comprehend the situation. Herlock Sholmes laughed and said:
“Wilson, you should conceal your astonishment at an incident
which is one of the most natural in the world.”
“Why do you not arrest him?” stammered Wilson.
 “Have you not observed, Wilson, that the gentleman is between
me and the door, and only a few steps from the door. By the time I
could move my little finger he would be outside.”
“Don’t let that make any difference,” said Lupin, who now walked
around the table and seated himself so that the Englishman was
between him and the door—thus placing himself at the mercy of the
foreigner.
Wilson looked at Sholmes to find out if he had the right to admire
this act of wanton courage. The Englishman’s face was
impenetrable; but, a moment later, he called:
“Waiter!”
When the waiter came he ordered soda, beer and whisky. The
treaty of peace was signed—until further orders. In a few moments
the four men were conversing in an apparently friendly manner.

Herlock Sholmes is a man such as you might meet every day in

the business world. He is about fifty years of age, and looks as if he
might have passed his life in an office, adding up columns of dull
figures or writing out formal statements of business accounts. There
was nothing to distinguish him from the average citizen of London,
except the appearance of his eyes, his terribly keen and penetrating
eyes.
But then he is Herlock Sholmes—which means that he is a
wonderful combination of intuition, observation, clairvoyance and
ingenuity. One could readily believe that nature had been pleased to
take the two most extraordinary detectives that the imagination of
man has hitherto conceived, the Dupin of Edgar Allen Poe and the
Lecoq of Emile Gaboriau, and, out of that material, constructed a
new detective, more extraordinary and supernatural than either of
them. And when a person reads the history of his exploits, which
have made him famous throughout the entire world, he asks himself
whether Herlock Sholmes is not a mythical personage, a fictitious
hero born in the brain of a great novelist—Conan Doyle, for instance.
When Arsène Lupin questioned him in regard to the length of his
sojourn in France he turned the conversation into its proper channel
by saying:
“That depends on you, monsieur.”
“Oh!” exclaimed Lupin, laughing, “if it depends on me you can
return to England to-night.”
“That is a little too soon, but I expect to return in the course of
eight or nine days—ten at the outside.”
“Are you in such a hurry?”
“I have many cases to attend to; such as the robbery of the
Anglo-Chinese Bank, the abduction of Lady Eccleston.... But, don’t
you think, Monsieur Lupin, that I can finish my business in Paris
within a week?”
“Certainly, if you confine your efforts to the case of the blue
diamond. It is, moreover, the length of time that I require to make
preparations for my safety in case the solution of that affair should
give you certain dangerous advantages over me.”
“And yet,” said the Englishman, “I expect to close the business in
eight or ten days.”
“And arrest me on the eleventh, perhaps?”
“No, the tenth is my limit.”
Lupin shook his head thoughtfully, as he said:
“That will be difficult—very difficult.”
“Difficult, perhaps, but possible, therefore certain—”
“Absolutely certain,” said Wilson, as if he had clearly worked out
the long series of operations which would conduct his collaborator to
the desired result.
“Of course,” said Herlock Sholmes, “I do not hold all the trump
cards, as these cases are already several months old, and I lack
certain information and clues upon which I am accustomed to base
my investigations.”
“Such as spots of mud and cigarette ashes,” said Wilson, with an
air of importance.
“In addition to the remarkable conclusions formed by Monsieur
Ganimard, I have obtained all the articles written on the subject, and
have formed a few deductions of my own.”
“Some ideas which were suggested to us by analysis or
hypothesis,” added Wilson, sententiously.
“I wish to enquire,” said Arsène Lupin, in that deferential tone
which he employed in speaking to Sholmes, “would I be indiscreet if
I were to ask you what opinion you have formed about the case?”
Really, it was a most exciting situation to see those two men
facing each other across the table, engaged in an earnest discussion
as if they were obliged to solve some abstruse problem or come to
an agreement upon some controverted fact. Wilson was in the
seventh heaven of delight. Herlock Sholmes filled his pipe slowly,
lighted it, and said:
“This affair is much simpler than it appeared to be at first sight.”
“Much simpler,” said Wilson, as a faithful echo.
“I say ‘this affair,’ for, in my opinion, there is only one,” said
Sholmes. “The death of the Baron d’Hautrec, the story of the ring,
and, let us not forget, the mystery of lottery ticket number 514, are
only different phases of what one might call the mystery of the
blonde Lady. Now, according to my view, it is simply a question of
discovering the bond that unites those three episodes in the same
story—the fact which proves the unity of the three events.
Ganimard, whose judgment is rather superficial, finds that unity in
the faculty of disappearance; that is, in the power of coming and
going unseen and unheard. That theory does not satisfy me.”
“Well, what is your idea?” asked Lupin.
“In my opinion,” said Sholmes, “the characteristic feature of the
three episodes is your design and purpose of leading the affair into a
certain channel previously chosen by you. It is, on your part, more
than a plan; it is a necessity, an indispensable condition of success.”
“Can you furnish any details of your theory?”
“Certainly. For example, from the beginning of your conflict with
Monsieur Gerbois, is it not evident that the apartment of Monsieur
Detinan is the place selected by you, the inevitable spot where all
the parties must meet? In your opinion, it was the only safe place,
and you arranged a rendezvous there, publicly, one might say, for
the blonde Lady and Mademoiselle Gerbois.”
“The professor’s daughter,” added Wilson.
“Now, let us consider the case of the blue diamond. Did you try to
appropriate it while the Baron d’Hautrec possessed it? No. But the
baron takes his brother’s house. Six months later we have the
intervention of Antoinette Bréhat and the first attempt. The diamond
escapes you, and the sale is widely advertised to take place at the
Drouot auction-rooms. Will it be a free and open sale? Is the richest
amateur sure to carry off the jewel? No. Just as the banker
Herschmann is on the point of buying the ring, a lady sends him a
letter of warning, and it is the Countess de Crozon, prepared and
influenced by the same lady, who becomes the purchaser of the
diamond. Will the ring disappear at once? No; you lack the
opportunity. Therefore, you must wait. At last the Countess goes to
her château. That is what you were waiting for. The ring disappears.”
“To reappear again in the tooth-powder of Herr Bleichen,”
remarked Lupin.
“Oh! such nonsense!” exclaimed Sholmes, striking the table with
his fist, “don’t tell me such a fairy tale. I am too old a fox to be led
away by a false scent.”
“What do you mean?”
“What do I mean?” said Sholmes, then paused a moment as if he
wished to arrange his effect. At last he said:
“The blue diamond that was found in the tooth-powder was false.
You kept the genuine stone.”
 Arsène Lupin remained silent for a moment; then, with his eyes
fixed on the Englishman, he replied, calmly:
“You are impertinent, monsieur.”
“Impertinent, indeed!” repeated Wilson, beaming with admiration.
“Yes,” said Lupin, “and, yet, to do you credit, you have thrown a
strong light on a very mysterious subject. Not a magistrate, not a
special reporter, who has been engaged on this case, has come so
near the truth. It is a marvellous display of intuition and logic.”
“Oh! a person has simply to use his brains,” said Herlock Sholmes,
nattered at the homage of the expert criminal.
“And so few have any brains to use,” replied Lupin. “And, now,
that the field of conjectures has been narrowed down, and the
rubbish cleared away——”
“Well, now, I have simply to discover why the three episodes were
enacted at 25 rue Clapeyron, 134 avenue Henri-Martin, and within
the walls of the Château de Crozon and my work will be finished.
What remains will be child’s play. Don’t you think so?”
“Yes, I think you are right.”
“In that case, Monsieur Lupin, am I wrong in saying that my
business will be finished in ten days?”
“In ten days you will know the whole truth,” said Lupin.
“And you will be arrested.”
“No.”
“No?”
“In order that I may be arrested there must occur such a series of
improbable and unexpected misfortunes that I cannot admit the
possibility of such an event.”
“We have a saying in England that ‘the unexpected always
happens.’”
They looked at each other for a moment calmly and fearlessly,
without any display of bravado or malice. They met as equals in a
contest of wit and skill. And this meeting was the formal crossing of
swords, preliminary to the duel.
“Ah!” exclaimed Lupin, “at last I shall have an adversary worthy of
the name—one whose defeat will be the proudest achievement in
my career.”
“Are you not afraid!” asked Wilson.
“Almost, Monsieur Wilson,” replied Lupin, rising from his chair,
“and the proof is that I am about to make a hasty retreat. Then, we
will say ten days, Monsieur Sholmes?”
“Yes, ten days. This is Sunday. A week from next Wednesday, at
eight o’clock in the evening, it will be all over.”
“And I shall be in prison?”
“No doubt of it.”
“Ha! not a pleasant outlook for a man who gets so much
enjoyment out of life as I do. No cares, a lively interest in the affairs
of the world, a justifiable contempt for the police, and the consoling
sympathy of numerous friends and admirers. And now, behold, all
that is about to be changed! It is the reverse side of the medal. After
sunshine comes the rain. It is no longer a laughing matter. Adieu!”
“Hurry up!” said Wilson, full of solicitude for a person in whom
Herlock Sholmes had inspired so much respect, “do not lose a
minute.”
“Not a minute, Monsieur Wilson; but I wish to express my
pleasure at having met you, and to tell you how much I envy the
master in having such a valuable assistant as you seem to be.”
Then, after they had courteously saluted each other, like
adversaries in a duel who entertain no feeling of malice but are
obliged to fight by force of circumstances, Lupin seized me by the
arm and drew me outside.
“What do you think of it, dear boy? The strange events of this
evening will form an interesting chapter in the memoirs you are now
preparing for me.”
 He closed the door of the restaurant behind us, and, after taking a
few steps, he stopped and said:
“Do you smoke?”
“No. Nor do you, it seems to me.”
“You are right, I don’t.”
He lighted a cigarette with a wax-match, which he shook several
times in an effort to extinguish it. But he threw away the cigarette
immediately, ran across the street, and joined two men who
emerged from the shadows as if called by a signal. He conversed
with them for a few minutes on the opposite sidewalk, and then
returned to me.
“I beg your pardon, but I fear that cursed Sholmes is going to give
me trouble. But, I assure you, he is not yet through with Arsène
Lupin. He will find out what kind of fuel I use to warm my blood.
And now—au revoir! The genial Wilson is right; there is not a
moment to lose.”
He walked away rapidly.
Thus ended the events of that exciting evening, or, at least, that
part of them in which I was a participant. Subsequently, during the
course of the evening, other stirring incidents occurred which have
come to my knowledge through the courtesy of other members of
that unique dinner-party.

At the very moment in which Lupin left me, Herlock Sholmes rose
from the table, and looked at his watch.
“Twenty minutes to nine. At nine o’clock I am to meet the Count
and Countess at the railway station.”
“Then, we must be off!” exclaimed Wilson, between two drinks of
whisky.
They left the restaurant.
 “Wilson, don’t look behind. We may be followed, and, in that case,
let us act as if we did not care. Wilson, I want your opinion: why
was Lupin in that restaurant?”
“To get something to eat,” replied Wilson, quickly.
“Wilson, I must congratulate you on the accuracy of your
deduction. I couldn’t have done better myself.”
Wilson blushed with pleasure, and Sholmes continued:
“To get something to eat. Very well, and, after that, probably, to
assure himself whether I am going to the Château de Crozon, as
announced by Ganimard in his interview. I must go in order not to
disappoint him. But, in order to gain time on him, I shall not go.”
“Ah!” said Wilson, nonplused.
“You, my friend, will walk down this street, take a carriage, two,
three carriages. Return later and get the valises that we left at the
station, and make for the Elysée-Palace at a galop.”
“And when I reach the Elysée-Palace?”
“Engage a room, go to sleep, and await my orders.”
Quite proud of the important rôle assigned to him, Wilson set out
to perform his task. Herlock Sholmes proceeded to the railway
station, bought a ticket, and repaired to the Amiens’ express in
which the Count and Countess de Crozon were already installed. He
bowed to them, lighted his pipe, and had a quiet smoke in the
corridor. The train started. Ten minutes later he took a seat beside
the Countess, and said to her:
“Have you the ring here, madame?”
“Yes.”
“Will you kindly let me see it?”
He took it, and examined it closely.
“Just as I suspected: it is a manufactured diamond.”
“A manufactured diamond?”
“Yes; a new process which consists in submitting diamond dust to
a tremendous heat until it melts and is then molded into a single
stone.”
“But my diamond is genuine.”
“Yes, your diamond is; but this is not yours.”
“Where is mine?”
“It is held by Arsène Lupin.”
“And this stone?”
“Was substituted for yours, and slipped into Herr Bleichen’s tooth-
powder, where it was afterwards found.”
“Then you think this is false?”
“Absolutely false.”
The Countess was overwhelmed with surprise and grief, while her
husband scrutinized the diamond with an incredulous air. Finally she
stammered:
“Is it possible? And why did they not merely steal it and be done
with it? And how did they steal it?”
“That is exactly what I am going to find out.”
“At the Château de Crozon?”
“No. I shall leave the train at Creil and return to Paris. It is there
the game between me and Arsène Lupin must be played. In fact, the
game has commenced already, and Lupin thinks I am on my way to
the château.”
“But—”
“What does it matter to you, madame? The essential thing is your
diamond, is it not?”
“Yes.”
“Well, don’t worry. I have just undertaken a much more difficult
task than that. You have my promise that I will restore the true
diamond to you within ten days.”
The train slackened its speed. He put the false diamond in his
pocket and opened the door. The Count cried out:
“That is the wrong side of the train. You are getting out on the
tracks.”
 “That is my intention. If Lupin has anyone on my track, he will
lose sight of me now. Adieu.”
An employee protested in vain. After the departure of the train,
the Englishman sought the station-master’s office. Forty minutes
later he leaped into a train that landed him in Paris shortly before
midnight. He ran across the platform, entered the lunch-room, made
his exit at another door, and jumped into a cab.
“Driver—rue Clapeyron.”
Having reached the conclusion that he was not followed, he
stopped the carriage at the end of the street, and proceeded to
make a careful examination of Monsieur Detinan’s house and the
two adjoining houses. He made measurements of certain distances
and entered the figures in his notebook.
“Driver—avenue Henri-Martin.”
At the corner of the avenue and the rue de la Pompe, he
dismissed the carriage, walked down the street to number 134, and
performed the same operations in front of the house of the late
Baron d’Hautrec and the two adjoining houses, measuring the width
of the respective façades and calculating the depth of the little
gardens that stood in front of them.
The avenue was deserted, and was very dark under its four rows
of trees, between which, at considerable intervals, a few gas-lamps
struggled in vain to light the deep shadows. One of them threw a
dim light over a portion of the house, and Sholmes perceived the
“To-let” sign posted on the gate, the neglected walks which encircled
the small lawn, and the large bare windows of the vacant house.
“I suppose,” he said to himself, “the house has been unoccupied
since the death of the baron.... Ah! if I could only get in and view
the scene of the murder!”
No sooner did the idea occur to him than he sought to put it in
execution. But how could he manage it? He could not climb over the
gate; it was too high. So he took from his pocket an electric lantern
and a skeleton key which he always carried. Then, to his great
surprise, he discovered that the gate was not locked; in fact, it was
open about three or four inches. He entered the garden, and was
careful to leave the gate as he had found it—partly open. But he had
not taken many steps from the gate when he stopped. He had seen
a light pass one of the windows on the second floor.
He saw the light pass a second window and a third, but he saw
nothing else, except a silhouette outlined on the walls of the rooms.
The light descended to the first floor, and, for a long time, wandered
from room to room.
“Who the deuce is walking, at one o’clock in the morning, through
the house in which the Baron d’Hautrec was killed?” Herlock
Sholmes asked himself, deeply interested.
There was only one way to find out, and that was to enter the
house himself. He did not hesitate, but started for the door of the
house. However, at the moment when he crossed the streak of
gaslight that came from the street-lamp, the man must have seen
him, for the light in the house was suddenly extinguished and
Herlock Sholmes did not see it again. Softly, he tried the door. It was
open, also. Hearing no sound, he advanced through the hallway,
encountered the foot of the stairs, and ascended to the first floor.
Here there was the same silence, the same darkness.
He entered, one of the rooms and approached a window through
which came a feeble light from the outside. On looking through the
window he saw the man, who had no doubt descended by another
stairway and escaped by another door. The man was threading his
way through the shrubbery which bordered the wall that separated
the two gardens.
“The deuce!” exclaimed Sholmes, “he is going to escape.”
He hastened down the stairs and leaped over the steps in his
eagerness to cut off the man’s retreat. But he did not see anyone,
and, owing to the darkness, it was several seconds before he was
able to distinguish a bulky form moving through the shrubbery. This
gave the Englishman food for reflection. Why had the man not made
his escape, which he could have done so easily? Had he remained in
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!

ebookname.com