100% found this document useful (4 votes)
55 views62 pages

Get Nature Inspired Optimization Techniques For Image Processing Applications Jude Hemanth Free All Chapters

Nature

Uploaded by

emefolocko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (4 votes)
55 views62 pages

Get Nature Inspired Optimization Techniques For Image Processing Applications Jude Hemanth Free All Chapters

Nature

Uploaded by

emefolocko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 62

Download the full version of the textbook now at textbookfull.

com

Nature Inspired Optimization Techniques for


Image Processing Applications Jude Hemanth

https://textbookfull.com/product/nature-inspired-
optimization-techniques-for-image-processing-
applications-jude-hemanth/

Explore and download more textbook at https://textbookfull.com


Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Nature Inspired Computing and Optimization Theory and


Applications 1st Edition Srikanta Patnaik

https://textbookfull.com/product/nature-inspired-computing-and-
optimization-theory-and-applications-1st-edition-srikanta-patnaik/

textbookfull.com

Human Behaviour Analysis Using Intelligent Systems D. Jude


Hemanth

https://textbookfull.com/product/human-behaviour-analysis-using-
intelligent-systems-d-jude-hemanth/

textbookfull.com

Image processing and GIS for remote sensing : techniques


and applications Second Edition Liu

https://textbookfull.com/product/image-processing-and-gis-for-remote-
sensing-techniques-and-applications-second-edition-liu/

textbookfull.com

Immigration and Housing Rents: Evidence from German


Reunification 1st Edition Kathleen Kürschner (Auth.)

https://textbookfull.com/product/immigration-and-housing-rents-
evidence-from-german-reunification-1st-edition-kathleen-kurschner-
auth/
textbookfull.com
A Country Nurse: From Wave Hill to rural Queensland and
almost everywhere in between Thea Hayes

https://textbookfull.com/product/a-country-nurse-from-wave-hill-to-
rural-queensland-and-almost-everywhere-in-between-thea-hayes/

textbookfull.com

Mastering UI Development With Unity An In Depth Guide to


Developing Engaging User Interfaces with Unity 5 Unity
2017 and Unity 2018 1st Edition Ashley Godbold
https://textbookfull.com/product/mastering-ui-development-with-unity-
an-in-depth-guide-to-developing-engaging-user-interfaces-with-
unity-5-unity-2017-and-unity-2018-1st-edition-ashley-godbold/
textbookfull.com

Babies in Groups 1st Edition Ben S. Bradley

https://textbookfull.com/product/babies-in-groups-1st-edition-ben-s-
bradley/

textbookfull.com

501 Spanish Verbs Christopher Kendris

https://textbookfull.com/product/501-spanish-verbs-christopher-
kendris/

textbookfull.com

Structural Information and Communication Complexity 20th


International Colloquium SIROCCO 2013 Ischia Italy July 1
3 2013 Revised Selected Papers 1st Edition Andrea Clementi
https://textbookfull.com/product/structural-information-and-
communication-complexity-20th-international-colloquium-
sirocco-2013-ischia-italy-july-1-3-2013-revised-selected-papers-1st-
edition-andrea-clementi/
textbookfull.com
Comptia A+ Core 1 (220-1001) and Core 2 (220-1002) Exam
Cram 2nd Edition David L. Prowse

https://textbookfull.com/product/comptia-a-core-1-220-1001-and-
core-2-220-1002-exam-cram-2nd-edition-david-l-prowse/

textbookfull.com
Intelligent Systems Reference Library 150

Jude Hemanth · Valentina Emilia Balas


Editors

Nature Inspired
Optimization
Techniques for
Image Processing
Applications
Intelligent Systems Reference Library

Volume 150

Series editors
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: [email protected]

Lakhmi C. Jain, Faculty of Engineering and Information Technology, Centre for


Artificial Intelligence, University of Technology, Sydney, NSW, Australia;
Faculty of Science, Technology and Mathematics, University of Canberra,
Canberra, ACT, Australia;
KES International, Shoreham-by-Sea, UK
e-mail: [email protected]; [email protected]
The aim of this series is to publish a Reference Library, including novel advances
and developments in all aspects of Intelligent Systems in an easily accessible and
well structured form. The series includes reference works, handbooks, compendia,
textbooks, well-structured monographs, dictionaries, and encyclopedias. It contains
well integrated knowledge and current information in the field of Intelligent
Systems. The series covers the theory, applications, and design methods of
Intelligent Systems. Virtually all disciplines such as engineering, computer science,
avionics, business, e-commerce, environment, healthcare, physics and life science
are included. The list of topics spans all the areas of modern intelligent systems
such as: Ambient intelligence, Computational intelligence, Social intelligence,
Computational neuroscience, Artificial life, Virtual society, Cognitive systems,
DNA and immunity-based systems, e-Learning and teaching, Human-centred
computing and Machine ethics, Intelligent control, Intelligent data analysis,
Knowledge-based paradigms, Knowledge management, Intelligent agents,
Intelligent decision making, Intelligent network security, Interactive entertainment,
Learning paradigms, Recommender systems, Robotics and Mechatronics including
human-machine teaming, Self-organizing and adaptive systems, Soft computing
including Neural systems, Fuzzy systems, Evolutionary computing and the Fusion
of these paradigms, Perception and Vision, Web intelligence and Multimedia.

More information about this series at http://www.springer.com/series/8578


Jude Hemanth Valentina Emilia Balas

Editors

Nature Inspired Optimization


Techniques for Image
Processing Applications

123
Editors
Jude Hemanth Valentina Emilia Balas
Department of Electronics and Department of Automation and Applied
Communication Engineering Informatics
Karunya University Aurel Vlaicu University of Arad
Coimbatore, Tamil Nadu, India Arad, Romania

ISSN 1868-4394 ISSN 1868-4408 (electronic)


Intelligent Systems Reference Library
ISBN 978-3-319-96001-2 ISBN 978-3-319-96002-9 (eBook)
https://doi.org/10.1007/978-3-319-96002-9

Library of Congress Control Number: 2018948603

© Springer International Publishing AG, part of Springer Nature 2019


This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Foreword

The active quest to endow machines with human abilities has been a feature of
modern times. The ultimate goal of creating an artificially intelligent and autono-
mous entity has been approached through many intermediate steps by providing
human-like functionality in a myriad of applications, including industrial automa-
tion, health care and security. A chief biological function that has been pursued is
that of analysing and understanding visual information. Advances in image pro-
cessing and computer vision have been adopted in a range of applications and have
transformed what is possible to be done automatically and without the need for
human visual intervention. In certain applications, machine capabilities have even
surpassed what humans can do. However, while in some of these limited cases they
have outstripped the human capabilities in terms of scale and speed, there are still
areas where humans have the edge and, therefore, the search for better approaches
and algorithms for image understanding continues.
At the same time, a better understanding of the emergence of biological systems,
including humans, has drawn the designers of machine vision systems to try to learn
from Nature. Through a very long process, spanning millennia, the Nature’s own
search for effective autonomous entities has resulted in efficient and effective
mechanisms for understanding and interacting with the world. Scientists and
designers are now learning from the fruits of Nature’s long labour to expedite the
development of artificial systems.
This volume brings together some of these naturally inspired approaches for
image understanding in one place and also provides a sample of the vast array of
applications to which they can be applied. For the reader new to these approaches, it
will provide a good starting point and for the more advanced algorithm designers, it
may suggest new ideas that they have not considered before.

v
vi Foreword

The deep and vast experience of Nature is a great resource for engineers and
designers in their quest for novel solutions to the current and emerging challenges
that face humanity. It is hoped that this book will contribute to this quest and
strengthens the case for the continued study of Nature in search of new insights.

Canterbury, UK Farzin Deravi, CEng, FIET


August 2018 Professor of Information Engineering
Head of School of Engineering & Digital Arts
University of Kent
Preface

This edited book is one of the significant contributions in the field of intelligent
systems for practical applications. This book is interdisciplinary with a wide cov-
erage of topics from nature-inspired optimization techniques and image processing
applications. The main objective of this book is to highlight the state-of-the-art
methods in these interdisciplinary areas to the researchers and academicians.
Variety of practical applications are covered in this book which can assist the
budding researcher to choose his own area of research. This book also covers
in-depth analysis of the methods which will attract high-end researchers to further
explore or innovate in these areas. In a nutshell, this book is a complete product for
usage by anyone working in the areas of intelligent systems. A brief introduction
about each chapter is as follows.
Chapter 1 illustrates the application of firefly optimization algorithm for brain
image analysis. Specifically, the methodology of CT and MRI brain image seg-
mentation is analysed in detail. Chapter 2 deals with image compression using bat
optimization techniques. An in-depth analysis of codebook generation for image
compression is analysed which will attract the readers. Chapter 3 deals with natural
language processing using particle swarm optimization methods. Few modified
swarm approaches are suggested in this chapter for efficient categorization of
alphabets in languages. The proposed approach is tested with Tamil language but
can be extended to different languages across the globe.
Chapter 4 covers the application of grey wolf optimization algorithm for image
steganography applications. Feature optimization for efficient data hiding is the
main objective of the work covered in this chapter. Literature survey is one critical
area of research which will attract several readers. With this idea, a detailed survey
on nature-inspired techniques for image processing applications is dealt in Chap. 5.
The application for ant colony optimization for visual cryptography is discussed in
Chap. 6. The primary focus of this work is image enhancement which can assist in
developing efficient cryptographic systems. Qualitative and quantitative analyses
are covered in this chapter which is more beneficial to the readers.

vii
viii Preface

The necessity of image analysis methods is significantly increasing in the area of


agriculture. The application of swarm intelligence techniques for detecting the
quality of crops via images is illustrated in Chap. 7. Analysing the quality of
different stages of wheat is the main focus of this chapter. Chapter 8 discusses the
various concepts of image preprocessing using cuckoo search optimization tech-
niques. Different types of input images are used in this chapter to validate the
proposed methodology. Automatic skin disease identification in mango fruits using
artificial bee colony algorithm is the focus of Chap. 9. The optimization algorithm is
used to select the optimal features for skin classification in this chapter.
Chapter 10 covers the different optimization techniques for fixing the structure
of the complex deep convolutional neural networks. An efficient architecture will
enhance the performance of any automated system. Chapter 11 deals with the
application of differential evolution method for quality enhancement in underwater
images. Chapter 12 covers the application of genetic algorithm for biometrics
application. Fetal biometrics-based abnormality detection is the prime focus of this
chapter.
We are grateful to the authors and reviewers for their excellent contributions for
making this book possible.
Our special thanks go to Janus Kacprzyk and Lakhmi C. Jain (Series Editors to
Intelligent Systems Reference Library) for the opportunity to organize this
guest-edited volume.
We are grateful to Springer, especially to Dr. Thomas Ditzinger (Senior Editor)
for the excellent collaboration.
We would like to express our gratitude and thanks to Handling Editor Ms.
Rajalakshmi Narayanan, Springer, Chennai and her team for their wholehearted
editorial support and assistance while preparing the manuscript.
This edited book covers the fundamental concepts and application areas in detail
which is one of the main advantages of this book. Being an interdisciplinary book,
we hope it will be useful to a wide variety of readers and will provide useful
information to professors, researchers and graduated, and all will find this collection
of papers inspiring, informative and useful.

Coimbatore, India Jude Hemanth


Arad, Romania Valentina Emilia Balas
August 2018
Contents

1 Firefly Optimization Based Improved Fuzzy Clustering


for CT/MR Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . ... 1
S. N. Kumar, A. Lenin Fred, H. Ajay Kumar and P. Sebastin
Varghese
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Fuzzy C Means Clustering . . . . . . . . . . . . . . . . . . . . . 5
1.2.3 Firefly Optimization Algorithm . . . . . . . . . . . . . . . . . . 8
1.2.4 Improved FCM-Firefly Optimization Segmentation
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Bat Optimization Based Vector Quantization Algorithm for
Medical Image Compression . . . . . . . . . . . . . . . . . . . . . . . . . ..... 29
A. Lenin Fred, S. N. Kumar, H. Ajay Kumar and W. Abisha
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.1 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.2 Overview of Image Compression . . . . . . . . . . . . . . . . 32
2.2.3 Vector Quantization Scheme . . . . . . . . . . . . . . . . . . . . 37
2.2.4 Linde Buzo Gray Algorithm . . . . . . . . . . . . . . . . . . . . 40
2.2.5 Bat Optimization Algorithm . . . . . . . . . . . . . . . . . . . . 42
2.2.6 Bat-VQ Image Compression Algorithm . . . . . . . . . . . . 45
2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

ix
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
x Contents

3 An Assertive Framework for Automatic Tamil Sign Language


Recognition System Using Computational Intelligence . . . . . . . ... 55
M. Krishnaveni, P. Subashini and T. T. Dhivyaprabha
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Proposed Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.1 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.2 Optimization Algorithms for Noise Removal . . . . . . . . 62
3.3.3 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.4 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.3.5 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.6 Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4 Improved Detection of Steganographic Algorithms in Spatial LSB
Stego Images Using Hybrid GRASP-BGWO Optimisation . . . . . .. 89
S. T. Veena, S. Arivazhagan and W. Sylvia Lilly Jebarani
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.2 Basics of Spatial LSB Algorithms . . . . . . . . . . . . . . . . . . . . . . 92
4.3 Proposed Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.1 Local Residue Pattern (LRP) . . . . . . . . . . . . . . . . . . . . 93
4.3.2 Local Distance Pattern (LDiP) . . . . . . . . . . . . . . . . . . . 96
4.4 Proposed Feature Selection Technique . . . . . . . . . . . . . . . . . . . 97
4.4.1 Binary Grey Wolf Optimisation (BGWO) . . . . . . . . . . 97
4.5 Experimental Results and Discussion . . . . . . . . . . . . . . . . . . . . 99
4.5.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5.2 Algorithm Detection Using Individual LRP and LDiP
Feature Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.5.3 Algorithm Detection Using Optimally Concatenated
LRP + LDiP Features . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.5.4 Algorithm Detection Using Optimised LRP + LDiP
Feature Using GB . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.5.5 Comparison with Existing Works . . . . . . . . . . . . . . . . 106
4.5.6 Algorithm Detection in Content Adaptive
Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5 Nature Inspired Optimization Techniques for Image
Processing—A Short Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
S. R. Jino Ramson, K. Lova Raju, S. Vishnu
and Theodoros Anagnostopoulos
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Contents xi

5.1.1 Nature Inspired Optimization Algorithms . . . . . . . . . . . 114


5.2 Evolutionary Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2.1 Classification of Evolutionary Algorithms . . . . . . . . . . 115
5.3 Swarm Intelligence Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 119
5.3.1 Gray Wolf Optimization (GWO) . . . . . . . . . . . . . . . . . 119
5.3.2 Bat-Algorithm (BA) . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.3.3 Ant Colony Optimization (ACO) . . . . . . . . . . . . . . . . . 123
5.3.4 Artificial Bee Colony Optimization (ABC) . . . . . . . . . . 126
5.3.5 Particle Swarm Optimization (PSO) . . . . . . . . . . . . . . . 127
5.3.6 Firefly Optimization (FFO) . . . . . . . . . . . . . . . . . . . . . 129
5.3.7 Cuckoo Search Algorithm (CS) . . . . . . . . . . . . . . . . . . 132
5.3.8 Elephant Herding Optimization (EHO) . . . . . . . . . . . . 134
5.3.9 Bumble Bees Mating Optimization (BBMO) . . . . . . . . 136
5.3.10 Lion Optimization Algorithm (LOA) . . . . . . . . . . . . . . 137
5.3.11 Water Wave Optimization (WWO) . . . . . . . . . . . . . . . 137
5.3.12 Chemical Reaction Optimization
Algorithm (CRO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.3.13 Plant Optimization Algorithm (POA) . . . . . . . . . . . . . . 140
5.3.14 The Raven Roosting Algorithm (RRO) . . . . . . . . . . . . 140
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6 Application of Ant Colony Optimization for Enhancement
of Visual Cryptography Images . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
G. Germine Mary and M. Mary Shanthi Rani
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.2 Review of Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.3 Image Enhancement of VC Shares Using ACO . . . . . . . . . . . . 150
6.3.1 Image Enhancement in VC Shares . . . . . . . . . . . . . . . . 150
6.3.2 Basics of ACO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.3.3 VC Share Enhancement Using ACO . . . . . . . . . . . . . . 152
6.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.4.1 Average Information Content (AIC) . . . . . . . . . . . . . . 155
6.4.2 Contrast Improvement Index (CII) . . . . . . . . . . . . . . . . 157
6.4.3 PSNR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.4.4 Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.4.5 Universal Image Quality Index (Q) . . . . . . . . . . . . . . . 158
6.4.6 Absolute Mean Brightness Error (AMBE) . . . . . . . . . . 158
6.4.7 Edge Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.4.8 Image Enhancement Factor (IEF) . . . . . . . . . . . . . . . . 161
6.4.9 Qualitative Analysis of the Proposed Method . . . . . . . . 161
6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
xii Contents

7 Plant Phenotyping Through Image Analysis Using Nature


Inspired Optimization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . 165
S. Lakshmi and R. Sivakumar
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.1.1 Wheat Production and Cultivation . . . . . . . . . . . . . . . . 166
7.1.2 Wheat Phenotyping . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.1.3 Nature Inspired Optimization Techniques . . . . . . . . . . . 172
7.1.4 Social Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
7.2 Proposed Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.3 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
8 Cuckoo Optimization Algorithm (COA) for Image Processing . . . . 189
Noor A. Jebril and Qasem Abu Al-Haija
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
8.2 Image Enhancement Functions . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.2.1 Image Transformation Formulas . . . . . . . . . . . . . . . . . 192
8.2.2 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
8.2.3 Parameter Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
8.3 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8.4 An Introduction to Cuckoo Search Optimization Algorithm . . . . 199
8.5 Image Enhancement via Cuckoo Search Methodology . . . . . . . 201
8.6 Pseudo-code of Cuckoo Search and Algorithm
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
8.7 MSE and PSNR Value Calculations . . . . . . . . . . . . . . . . . . . . . 208
8.8 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
9 Artificial Bee Colony Based Feature Selection for Automatic Skin
Disease Identification of Mango Fruit . . . . . . . . . . . . . . . . . . . . . . . 215
A. Diana Andrushia and A. Trephena Patricia
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
9.2 Backgrounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.2.1 Fruit Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.2.2 Classification Methods . . . . . . . . . . . . . . . . . . . . . . . . 219
9.3 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
9.3.1 Input Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
9.3.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
9.3.3 Feature Extraction—Color, Shape,
Texture Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
9.3.4 Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
9.3.5 Support Vector Machine . . . . . . . . . . . . . . . . . . . . . . . 224
Contents xiii

9.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226


9.4.1 Background Removal . . . . . . . . . . . . . . . . . . . . . . . . . 226
9.4.2 ABC Based Feature Selection . . . . . . . . . . . . . . . . . . . 228
9.4.3 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 229
9.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
10 Analyzing the Effect of Optimization Strategies in Deep
Convolutional Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
S. Akila Agnes and J. Anitha
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
10.2 CNN Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
10.2.1 Convolution Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
10.2.2 Activation Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
10.2.3 Pooling Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
10.2.4 Fully Connected Layer . . . . . . . . . . . . . . . . . . . . . . . . 240
10.3 Proposed DCNN Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 241
10.3.1 Dropout Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
10.3.2 Batch Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . 242
10.3.3 Optimizing Gradient Descendant with Various
Optimizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
10.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
10.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
11 A Novel Underwater Image Enhancement Approach with
Wavelet Transform Supported by Differential Evolution
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Gur Emre Guraksin, Omer Deperlioglu and Utku Kose
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
11.2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
11.2.1 Wavelet Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
11.2.2 Differential Evolution Algorithm . . . . . . . . . . . . . . . . . 260
11.2.3 Contrast Adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . 261
11.2.4 Homomorphic Filter . . . . . . . . . . . . . . . . . . . . . . . . . . 261
11.2.5 Unsharp Masking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
11.3 Solution of the Proposed Image Enhancement Approach . . . . . . 264
11.4 Applications with the Proposed Approach . . . . . . . . . . . . . . . . 267
11.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
11.6 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . 274
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
xiv Contents

12 Feature Selection in Fetal Biometrics for Abnormality Detection


in Ultrasound Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
R. Ramya, K. Srinivasan, B. Sharmila and K. Priya Dharshini
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
12.2 Steps Involved in Processing of Fetal Image . . . . . . . . . . . . . . . 281
12.2.1 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
12.2.2 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
12.2.3 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
12.2.4 Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.3 Experimental Results and Discussion . . . . . . . . . . . . . . . . . . . . 289
12.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
Chapter 1
Firefly Optimization Based Improved
Fuzzy Clustering for CT/MR Image
Segmentation

S. N. Kumar, A. Lenin Fred, H. Ajay Kumar


and P. Sebastin Varghese

Abstract The segmentation is the process of extraction of the desired region of


interest. In medical images, the anatomical organs and anomalies like a tumor,
cysts, etc. are of importance for the diagnosis of diseases by physicians for tele-
medicine applications. The thresholding, region growing, and edge detection are
termed as classical segmentation algorithms. Clustering is an unsupervised learning
technique to group similar data points and fuzzy partitioning merges similar pixels
based on the fuzzy membership value. The classical FCM algorithm lacks sensi-
tivity in the cluster centroid initialization and often gets trapped in local minima.
The optimization algorithm gains its importance in cluster centroids initialization,
thereby improving the efficiency of FCM algorithm. In this work, firefly opti-
mization is coupled with FCM algorithm for CT/MR medical image segmentation.
Fireflies are insects having a natural capacity to illumine in dark with glowing and
flickering lights and firefly optimization algorithm was modeled based on its bio-
logical traits. The preprocessing stage comprises of artifacts removal and denoising
by Nonlinear Tensor Diffusion (NLTD) filter. The computation time was minimized
by reducing the total pixels count for the processing. The Firefly optimization, when
coupled with FCM, generates satisfactory results inconsistent with FCM when
coupled with Cuckoo, Artificial Bee Colony, and Simulated annealing algorithms.
The cluster validity performance metrics are used for the determination of optimum
number of clusters. The algorithms are developed in Matlab 2010a and tested on
real-time abdomen datasets.

S. N. Kumar (&)
Sathyabama Institute of Science and Technology, Chennai, India
e-mail: [email protected]
A. Lenin Fred  H. Ajay Kumar
Mar Ephraem College of Engineering and Technology, Elavuvilai, India
e-mail: [email protected]
H. Ajay Kumar
e-mail: [email protected]
P. Sebastin Varghese
Metro Scans and Research Laboratory, Trivandrum, India
e-mail: [email protected]

© Springer International Publishing AG, part of Springer Nature 2019 1


J. Hemanth and V. E. Balas (eds.), Nature Inspired Optimization Techniques
for Image Processing Applications, Intelligent Systems Reference Library 150,
https://doi.org/10.1007/978-3-319-96002-9_1
2 S. N. Kumar et al.

 
Keywords Unsupervised learning Clustering Fuzzy C means

FCM-firefly algorithm FCM-artificial bee colony algorithm FCM-cuckoo
algorithm

1.1 Introduction

Medical image processing refers to the application of computer-aided algorithms


for the extraction of anatomical organs and analysis of anomalies like a tumor, cyst,
etc. The various steps in image processing are restoration, enhancement, segmen-
tation, classification, and compression. The segmentation can be defined as the
extraction of Region of Interest (ROI). The Computer Tomography (CT), Magnetic
Resonance Imaging (MRI), Ultrasound and Positron Emission Tomography
(PET) are the widely used medical imaging modalities for the disease diagnosis.
The choice of segmentation algorithm depends on the medical imaging modality
and its characteristics.
The CT images, in general, are corrupted by Gaussian noise and its distribution
is as follows.

1 2
pðzÞ ¼ pffiffiffi eðxlÞ =2r
2

2pr

where x represents random variable normally distributed with mean l and standard
deviation r.
The MR images are corrupted by rician noise, artifacts and intensity inhomo-
geneity due to the non-uniform response of RF coil. The rician noise distribution is
as follows
 2 
z z þ I 2  za 
pðzÞ ¼ 2 exp  B 2
r 2r2 r

where, I is the true intensity value, r is the standard deviation of the noise, and B is
the modified zeroth order Bessel function.
The Ultrasound images, in general, are corrupted by speckle noise and its dis-
tribution is as follows.
 
gc1 g
FðxÞ ¼ ea
ðc  1Þ!ac

where, a is the variance, c is the shape parameter of gamma distribution and g is the
gray level.
Prior to segmentation, the preprocessing was performed by appropriate filtering
technique; Filter selection is based on the medical imaging modality and noise
characteristics. The role of preprocessing is inevitable in signal and image
1 Firefly Optimization Based Improved Fuzzy Clustering … 3

Segmentation
Algorithms

First Generation : Second Generation: Third Generation :


Thresholding Deformable, Classifier,
Clustering, Watershed, Graph Guided,
Region Growing
Markov Random Atlas , Hybrid
Edge based methods Field techniques approaches

Fig. 1.1 Classification of segmentation algorithms

processing for subsequent operations like segmentation and classification [1, 2].
The segmentation algorithms can be categorized based on the generation of evo-
lution and are depicted in Fig. 1.1.
Image segmentation is the process of grouping the pixels of an image to form
meaningful regions. Medical image segmentation is the visualization of the region
of interest such as anatomical structures and anomalies like tumor, cyst, etc. for
medical applications such as diagnostics, therapeutic planning, and guidance. Lay
Khoon Lee et al. performed a review on different types of segmentation algorithms
for medical imaging modalities like X-ray, CT, MRI, 3D MRI and Ultrasound [3].
Similarly, S. N. Kumar et al. performed a detailed study on the different generation
of the medical image segmentation techniques; qualitative and quantitative analysis
was performed for the widely used medical image segmentation algorithms [4].
Neeraj Sharma et al. state the necessity of automated medical image segmentation
technique in diagnosis, and radiotherapy planning in medical images and also
explained the limitations of the existing segmentation algorithms [5]. The thresh-
olding is a simple and classical technique that separates the foreground and
background regions in an image based on the threshold value. The multilevel
thresholding eliminates the discrepancy of the bi-level thresholding that uses a
single threshold value. The optimization techniques when employed in the multi-
level thresholding yield efficient results, since it provides the proper choice of
threshold values. The 3D Otsu thresholding was found to be efficient for MR brain
images; better results were produced than bi-level and multithresholding techniques
[6]. Among the region based approaches, the classical region growing is the
semi-automatic segmentation technique that relies on the seed point selection [6].
The multiple-seed point based region growing for brain segmentation was found to
be effective on a multi-core CPU computer [7]. The manual seed point selection can
be replaced by the deployment of the optimization algorithm for yielding efficient
results [8]. The edge detection traces the boundary of objects in an image and
among the classical edge detector, canny produces better results [9]. The Markov
basics and Laplace filter were coupled to form an edge detection model that gives
better results for medical images than the classical techniques [10]. The teaching
4 S. N. Kumar et al.

learning-based optimization was found to be effective for medical image edge


detection than the classical edge detectors [11]. The interactive medical image
segmentation algorithms are discussed in [12]. J. Senthilnath et al. did a perfor-
mance study on nature-inspired firefly optimization algorithm in the thirteen
benchmark classification datasets [13]. Superior results were produced when
compared with classical techniques like Particle Swarm Optimization (PSO), Bayes
net, Multilayer Perceptron, Radial Basis Function Neural Network, KStar, Bagging,
MultiBoost, Naive Bayes Tree, Ripple Down Rule, Voting Feature Interval.
Iztok Fister et al. made a detailed analysis of the types of firefly algorithm for
engineering applications in solving the real world challenges [14]. Hui Wang et al.
proposed a modification in the parameter of classical firefly algorithm to reduce the
complexity of the algorithm [15]. The proposed adaptive firefly algorithm generates
better solution when compared with standard Firefly Algorithm, Variable step size
Firefly Algorithm (VSSFA), Wise step strategy Firefly Algorithm (WSSFA),
Memetic Firefly Algorithm (MFA), Firefly Algorithm with chaos and Firefly
Algorithm with random attraction. Mutasem K et al. proposed a hybrid technique
comprising of the Fuzzy C-Means algorithm with Firefly algorithm for the seg-
mentation of brain tumor [16]. The experimental analysis was carried out on 181
brain images obtained from brain-web Simulated Brain Database (SBD) repository;
robust results were produced when compared with Dynamic clustering algorithm
based on the hybridization of Harmony Search and Fuzzy Variable String Length
Genetic Point symmetry techniques. K. Vennila et al. proposed multilevel Otsu
image segmentation based on Firefly optimization and good results were obtained
in terms of PSNR, computation cost and mean value when compared with
Darwinian Particle Swarm Optimization [17]. Cholavendhan Selvaraj et al. made a
detailed survey of the bio-inspired optimization algorithms such as Ant Colony
Optimization, Particle Swarm optimization, Artificial Bee Colony algorithm and
their hybridizations [18].
The summarization of results reflects the status of the optimization techniques in
solving the wide range of engineering problems. In the medical image processing,
the FCM plays a major role in the clustering and classification of the image for the
analysis, diagnosis, and recognition of anomalies [19]. Janmenjoy Nayak et al.
performed a survey on major modification and advancement in the classical FCM
algorithm and their applications towards the image analysis [20]. Chih Chin Lai
et al. proposed a hierarchical evolutionary algorithm based on genetic algorithm for
the segmentation of skull images which enhances the diagnostic efficiency than the
dynamic thresholding, Competitive Hopfield Neural Networks (CHNN), K-Means
and Fuzzy C-Means algorithms [21].
Emrah Hancer et al. proposed a methodology for the segmentation of brain
tumor in the MRI images by using artificial bee colony algorithm. Efficient results
were produced when compared with K-Means, FCM, and Genetic Algorithm based
image segmentation techniques [22]. The FCM, when coupled with PSO was found
to be effective for the segmentation of noisy images when compared with K-means,
Enhanced FCM, and Fast Global Fuzzy Clustering techniques [23].
1 Firefly Optimization Based Improved Fuzzy Clustering … 5

The Convolution Neural Network (CNN) was employed for the automatic
segmentation of MR brain images, multiple convolution kernels of varying size was
used for the generation of accurate results [24]. The CNN with multiple kernels of
smaller size was used for the efficient brain tumor segmentation in MR images [25].
The Deep Learning Neural Network (DLNN) gains its importance in attenuation
correction of PET/MR images [26]. The DLNN along with deformable model was
proposed for the automatic segmentation of left ventricle in cardiac MR images
[27]. The Deep Convolution Neural Network (DCNN) along with the 3D
deformable model generates good segmentation results for the extraction of tissues
in musculoskeletal MR images [28]. Vijay Badrinarayanan et al. proposed SegNet,
a novel DCNN architecture for semantic pixel-wise segmentation [29]. In this
chapter, firefly optimization algorithm was coupled with FCM for CT/MR medical
image segmentation. The preprocessing stage comprises of artifacts removal and
denoising by Non-Linear Tensor Diffusion (NLTD) filter. The computation com-
plexity of the algorithm was minimized by sampling the total pixel count for
manipulation. The Cluster Validity Indexes (CVI’s) are used for the validation of
results to determine the optimum number of clusters.

1.2 Materials and Methods

1.2.1 Data Acquisition

The real-time abdomen CT data sets are used in this work for the analysis of
algorithms. The images are acquired from Optima CT machine with a slice
thickness of 3 mm. The images in DICOM format with a size of 512  512 are
used in this work. The Metro Scans and Research Laboratory approved the study of
human datasets for research purpose. The five abdomen CT data sets, each com-
prising of 200 slices are used in this work. The results of typical slice from each
dataset are depicted here.

1.2.2 Fuzzy C Means Clustering

In this chapter, the Fuzzy c-means Clustering algorithm coupled with optimization
technique was proposed for the segmentation of medical images. In the perspective
of image processing, clustering is defined as the grouping of pixels into a cluster
which is similar between them, while dissimilar pixels belong to other clusters. The
concept of clustering is depicted below in Fig. 1.2. The clustering algorithms can be
classified into two groups; Supervised and Unsupervised. The requirement of prior
knowledge termed as training samples is the key concept of the supervised clas-
sifier. Artificial Neural Network (ANN), Naive Bayes Classifier, and Support
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
6 S. N. Kumar et al.

Fig. 1.2 Principle of clustering

Vector Machine are some of the widely used supervised algorithms. The unsu-
pervised technique doesn’t need any prior information and is particularly well
suited for huge unlabeled datasets. The unsupervised clustering techniques can be
further classified into two categories; hierarchical and partitional. The role of par-
titional clustering is prominent in image analysis and pattern recognition. The
K-means and Fuzzy c-means (FCM) are well-known partition clustering algo-
rithms. The K- means clustering is termed as Crisp (hard) since the objects are
assigned to only one cluster. The FCM clustering is termed as soft (Fuzzy) since an
object can be accommodated in more than one cluster based on the fuzzy mem-
bership value.
The FCM overcomes the issues of classical K-means clustering; since the data
can belongs to more than one cluster. The FCM was developed by Dunn [30] and
modified by Hathaway and Bezdek [31] which was widely used for pattern clas-
sification. FCM is an unsupervised algorithm based on the minimization of the
objective function.

N X
X C  2
Jm ¼ Uijm yi  cj  ; 1  f \1
i¼1 j¼1

The pixels are grouped into clusters in such a manner that, the intracluster
similarity is maximized and the intracluster similarity is minimized.
The fuzzy partition represents the fuzzy membership matrix of the pixel in the
cluster. The parameter Uij represents the fuzzy membership of the ith object (pixel)
in the jth cluster. The parameter ‘f’ depicts weighting exponent that determines the
degree of fuzziness for the fuzzy membership function. The fuzzy classification is
based on the iterative optimization of the objective function depicted above with the
updation of membership function uij and the cluster center cj as follows.
1 Firefly Optimization Based Improved Fuzzy Clustering … 7

1
Uij ¼  f 1
2
PC kyi cj k
K¼1 kyj ck k
PN
i¼1 Uijf  yi
c j ¼ PN
i¼1 Uijf
n o
 ðk þ 1Þ ðkÞ 
The iterative calculation is terminated, when maxij uij  uij  \d, where
d is a termination criterion between 0 and 1, and k represents the iteration count.
The convergence of the algorithm occurs when the objective function (Jm) attains
local minima or saddle point.
The steps in FCM clustering algorithm are summarized as follows

1. Initialise U ¼ ½Uij  matrix; U ð0Þ


2. At kth step: Calculate the cluster center vector CðkÞ ¼ cj with U ðkÞ
PM
i¼1 Uijm  xi
c j ¼ PM
i¼1 Uijm

3. Update U ðkÞ ; U ðk þ 1Þ

1
Uij ¼  m1
2
PC kxi cj k
K¼1 kxj ck k
 
4. If U ðk þ 1Þ  U ðkÞ \d; then Stop; otherwise return to step 2:

The operating principle of FCM is based on the fact that, the minimization of the
objective function ends up with the solution. In many real-time cases, classical
FCM stuck into local minima. The optimization algorithm can be employed to
achieve global minima. The parameter selection is vital for optimization algorithms
and it influences the performance of the algorithm to maximize or minimize the
objective function subjecting to certain constraints. The cluster centers are ran-
domly initialized by classical FCM, hence the optimization based clustering solves
this problem. The cluster centers generated by the optimization technique is utilized
by the FCM for image segmentation. The pixels in the image are mapped into the
8 S. N. Kumar et al.

particular cluster based on similarity and distance. The initialization of the cluster
centers by optimization improves the performance in terms of the convergence rate,
computation complexity, and segmentation accuracy.

1.2.3 Firefly Optimization Algorithm

In this chapter, the performance of firefly optimization in the FCM algorithm was
analyzed for the estimation of optimal cluster center values for image segmentation.
The biological trait of the firefly is the motivation for Yang [32] to propose an
optimization technique. The rhythmic flashes generated by the firefly was used as a
mode of communication between them to search for prey and for mating. More than
2000 species of fireflies are there in the world and they have natural characteristics
to create illumination in the dark with flickering and glowing lights. Fister et al.
found that the attraction capacity of the fireflies is proportional to the brightness
[14]. The fireflies tend to move towards ones which emits a brighter light.
The population-based firefly algorithm was found to generate a global optimal
solution for many engineering problems. The biological chemical substance luci-
ferin present in the body of the fireflies was responsible for flashing the light. The
intensity of light emitted is directly proportional to the discharge of luciferin. The
degree of attraction tends to decrease as the distance between the fireflies increases.
If any firefly fails to discover another firefly which is brighter than itself, it will
travel arbitrarily. The optimization algorithm when employed for clustering appli-
cations, cluster centers are the decision variables and the objective function is
associated with the euclidean distance. Based on the objective function, initially, all
the fireflies will be spread randomly over the search space.
The two stages of firefly algorithm are summarized as follows:
The first stage is based on the difference in the intensity that is associated with
the objective function values. Depending on the nature of the problem that requires
maximization/minimization, a firefly with higher/lower intensity will entice another
individual with higher/lower intensity.
Consider that there are n swarms (fireflies), where Yi signifies the solution of a
firefly i. The fitness value is expressed by f ðYi Þ moreover the current position I of
the fitness value f ðyÞ is estimated by the brightness of a firefly [32].

Ii ¼ f ðyi Þ; 1in

The second stage is the movement towards the firefly with high brightness
intensity. The attraction factor of the firefly is represented by b that indicates the
attraction power of firefly in the swarm and it changes with distance ðRij Þ between
two fireflies i and j at positions Yi and Yj respectively.
1 Firefly Optimization Based Improved Fuzzy Clustering … 9

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  Xd
Rij ¼ Yi  Yj  ¼
2
k¼1
Yik  Yjk

The attraction function bðRÞ of the firefly is expressed as follows.

bðRÞ ¼ b0 ecR
2

where b0 is the attraction function value for R ¼ 0; c is the coefficient of ingestion


of light.
The pseudo code for firefly optimization algorithm is as follows

Define objective function f(Y), Y=[Y1,Y2,Y3,--------Yd]

Generate initial population of fireflies Yi =[1,2,3-----n]

Estimate the light intensity of firefly Ii using the objective function f(Y)

Define light absorption coefficient(

While (t<max generation )

for i=1:n //all n fireflies

for j=1: n // all n fireflies

if (Ij > Ii)

Move firefly i towards j in d dimensions.

end if

// Attraction capacity changes with distance

//Validate new solutions and update light intensity

end for j

end for i

Estimate the current best by ranking the fireflies

end while

The motion of a firefly ‘i’ from the position Yi which is attracted towards another
brighter firefly ‘j’ at position Yj is expressed as follows
10 S. N. Kumar et al.

 
1
Yi ðt þ 1Þ ¼ Yi ðtÞ þ bðRÞ Yi  Yj þ a rand 
2
 
cR2 1
Yi ðt þ 1Þ ¼ Yi ðtÞ þ b0 e Yi  Yj þ a rand 
2

where a depicts the maximum radius of the random step. The term rand represents
randomization parameter uniformly distributed between 0 and 1.
There are two special cases
Case i: r ¼ 0, then b ¼ b0 e0 ¼ b0 , The air is absolutely clear with no light dis-
persion. The fireflies can see each other; exploration and exploitation is out of
balance.
Case ii: r ¼ 1, then b ¼ I0 e1d ¼ 0, The air is foggy with extreme light dis-
2

persion. The fireflies can’t see each other; exploration and exploitation is out of
balance.

1.2.4 Improved FCM-Firefly Optimization Segmentation


Algorithm

The FCM clustering algorithm proposed here comprises of two stages. In the first
stage, firefly optimization is employed to determine the near-optimal cluster centers.
In the second stage, the cluster centers are used for the initialization of FCM
algorithm. The firefly optimization algorithm makes the clustering an effective tool
for medical image segmentation by eliminating the problem of stucking at local
minima. The firefly optimization is a swarm intelligence based algorithm and hence
it mimics its advantages.
The solution vector is expressed as follows
 
V1 V2 V3

S1 ; S2 ; . . .; Si . . .Sd S1 ; S2 ; . . .; Si . . .Sd S1 ; S2 ; . . .; Si . . .Sd

where Si represents characteristics in numerical form such that Si € S. The ‘S’ depicts
the array representing pixel attribute. Each cluster center Vi is represented by d
numerical features ðS1 ; S2 ; . . .Sd Þ. Each solution vector is of the size (c  d), where c
indicates given number of clusters and d represents the features of the dataset.
For the delineation of anomalies like tumor or cyst or anatomical organs, each
pixel in the image is mapped into the clustering sector. The cluster centers are
randomly initialized from the image pixel gray values with the randomly initialized
solution vector, the fitness value is determined by the objective function. The
solution vector is then rearranged based on the decreasing order of the objective
function value. The firefly optimization determines near optimal cluster centers
thereby ensuring global minima for FCM algorithm and hence eliminates the
Random documents with unrelated
content Scribd suggests to you:
life, to the shade of the monastery, and now she was to spend the
last fifteen years of her life on the imperial throne. She was of
sterner stuff than Zoe, and the Senators must have concluded that
she alone could check the audacity of the low-born Paphlagonians.
This does not in itself argue any great strength of character in
Theodora. We must remember that there was always a party of
ambitious eunuchs or statesmen behind each of the names that is
put forward by the historian.
When the news of this decision reached Michael, and the crowd
stormed angrily at the gates of the palace, he sent an officer on a
swift vessel to the Princes’ Islands for Zoe. In the palace she was
quickly stripped of her nun’s robe, and clothed in her former
garments. It is clear that Michael’s uncle, Constantine, who was not
without ability, directed the campaign in the palace. Michael was
advised to take Zoe with him into the imperial lodge overlooking the
Hippodrome and show the citizens, who had gathered in the
enclosure, that all was well. The only reply he got was a shower of
stones, arrows and epithets, and, as the chroniclers remark, the
young lion became at once a timid hare, and proposed to run for
shelter to the monastery at Studion, on the Asiatic side. His uncle
prevented him, however, and marshalled the guards in the fore part
of the palace. The battle which followed ended in a complete victory
for the people. Constantine and Michael fled across the water to
Studion, in the early morning of Wednesday in Holy Week, and the
new Empress Theodora was conducted into the palace over the
corpses of some three thousand of the combatants.
The royal sisters, it will be understood, did not fly into each
other’s arms. Theodora had to thank Zoe for eleven years’
confinement, and Zoe herself was very reluctant to share her power
with her younger sister. However, a formal reconciliation was
arranged by the Senators, and the two Empresses sat side by side to
receive the homage of the leading citizens and decide what was to
be done with the late Emperor and his uncle. If there were any who
wondered in what spirit Theodora would wield her power after a
decade of religious life, they were not left long in doubt. Zoe asked
what the will of her advisers was in regard to the fugitives, and such
cries as “Out with their eyes!” and “Crucify them!” rang furiously
through the chamber. Zoe recoiled and pleaded for leniency, but
Theodora, a much better speaker than her sister, sternly ordered the
prefect to see that their eyes were put out. A great crowd crossed
the sea with the officers, and saw Michael, who had hidden under
the altar, and his more stoical uncle dragged from the chapel. The
same crowd had applauded Michael in his procession hardly a week
before; now they stood by with wild delight to see the brutal
sentence carried out. It was 21st April: Michael the Caulker had
reigned for four months.
For a few weeks the imperial sisters ruled their kingdom in
complete harmony and with exemplary zeal. M. Diehl, too lightly
following the censorious Psellus, rates the intelligence and character
of both at a very low level, but that estimate is hardly supported by
the facts. Few Emperors had dared to attack the administrative
corruption of the Empire as Zoe and Theodora attacked it in the first
freshness of their power, and as we have every reason to believe
that they would have continued to attack it. For centuries the State
had been the easy prey of ambitious eunuchs at Court and corrupt
officials in the provinces. Zoe and Theodora issued decrees to the
effect that all injustice must cease and that the law must be
administered with equity. They themselves sat on the highest
tribunal of the city to hear cases, and the sale of offices was strictly
prohibited. The accounts of the late chief minister were examined,
and Constantine, eyeless and shaven, was brought from his
monastery to explain the enormous deficiency. The power of his
family was broken for ever, and the miserable man disclosed that
5300 pounds of gold (nearly a quarter of a million sterling) was
hidden in a cistern in his house. Legates and petitions were heard
with dignity by the royal sisters, and it must have seemed to many
that the Empire had, by this singular adventure, obtained juster and
finer rulers than it had known for many a century. We cannot
discriminate in the joint public action of the sisters, but it is clear
that the strong will and intelligence of Theodora were the chief
power of the administration. How drastically the Empire needed such
a purification may be gathered from the fact that, when the
patriarch Alexis died in the following year, a secret and dishonest
hoard of gold, amounting to more than £100,000, was discovered in
his palace.
This brilliant example of feminine rule might have been expected
to disarm the old Byzantine prejudice against women, but prejudices
of that nature are too deeply rooted to be displaced by facts. The
cry was raised that an Emperor was needed, and Zoe once more
expressed her willingness to marry. The careful chronicler tells us
that her conduct was not necessarily inspired by a carnal feeling—
she was now sixty-two years old—but that she may have feared that
Theodora and her ministers wished to dislodge her. Her age, no less
than the remarkable conditions of her third and last marriage, will
easily persuade us that the motive was political. There were those
who said that, as Theodora had been the chief agent in expelling
Michael, the throne belonged to her alone, and Zoe sought an ally.
The first noble chosen by her was Constantine Delassenus, who had
almost obtained her hand and the throne fourteen years before. But
Constantine, when he was invited to the Court for inspection, proved
so brusque and independent that he was again dismissed. Her next
choice was Constantine Catepano, a handsome officer of the palace,
with whom, in spite of her age, the gossips of the Court already
connected Zoe somewhat too intimately. Constantine, however, had
a wife living, and this lady is said to have poisoned him as soon as
she heard of the proposal to divorce her.
If we may believe the gossipy chronicles, Zoe met the
disappointment with tranquillity, as she had another lover among the
officials of the palace. Constantine Monomachos, a very handsome
and distinguished and dissolute noble, had been exiled from Court to
Mitylene by Michael IV. on the suspicion of intimacy with Zoe, and
had for some years gilded the hours of his distant exile with the
enjoyment of letters, the pleasures of the table and the affection of
a pretty and devoted cousin. When his second wife had died, he had
obeyed the injunction of the Church to refrain from a third marriage
and had been content with the free companionship of the beautiful
Sclerena, a sister of the distinguished noble Romanus Sclerus—a
member, that is to say, of one of the proudest Byzantine families.
She had followed her lover to Lesbos, used her fortune to mitigate
the harshness of his exile, and was living with him at the time when
Zoe recalled him to Court. “Handsome as Achilles,” uniting a
prodigious strength with a singular delicacy and elegance of
appearance, equally devoted to the robust pleasures of the chase
and the enervating delights of love, Constantine Monomachos at
once returned to his place in the heart of the ageing Empress, and
was invited to wed her. He is said to have stipulated beforehand that
the fair Sclerena should be allowed to come to Constantinople, and
Zoe genially consented. They were married, and Zoe entered upon
the last and strangest part of her strange career.
While the sexless Theodora continued to rule the Empire and put
out the eyes of her enemies, while Constantine revelled in the new
and more exquisite luxuries of his position, Zoe seems quietly to
have enjoyed the secure and restful days which her marriage
obtained for her. She still, with her maids, compounded and distilled
the perfumes which were almost her one luxury, but she now paid a
scrupulous attention to her devotions and burned much incense
before the icons. Sclerena at first dwelt apart, and Constantine set
about building a magnificent palace for her, thinly veiling his liaison
with the pretence of going daily to see the progress of the works. As
the citizens smiled at the connexion, and Zoe seemed to be piously
indifferent to it, he became bolder and asked Zoe to allow him to
bring Sclerena to live in the palace. Again Zoe consented, and the
ménage à trois was maintained in the most pleasant harmony. She
gave Sclerena the title of Empress, embraced her, when they met,
with entire goodwill, and showed her such consideration that she
never visited her husband without first ascertaining if he was
disengaged. Constantine occupied the central part of the palace, and
his wife and mistress had apartments on each side.
Although Zoe now approached her seventieth year, she still
retained the freshness of her complexion and had no wrinkles.
Psellus says that a stranger would have been sure that she was still
a young woman. She shared the pleasures of the gay Court, and
made no protest against the frivolous Constantine emptying the
treasury on his mistress. If we may believe implicitly all the details
given by Psellus, there was little delicacy in the fun which enlivened
the gardens or halls—for Zoe disliked the open air—of the sacred
domain. Music and skilful dancing were too fine for his appreciation.
He liked the broader merriment of mimes, and took especial
pleasure in imitations of stammering. His chief entertainers would go
so far as to represent, pantomimically, the chaste Theodora lying
abed in child-birth, and Theodora herself joined in the loud laughter
of Constantine as the man imitated the shrieks which befitted such
an occasion. The months passed very merrily, and the treasury
emptied.
And as the treasury emptied, and the citizens saw their funds
passing into the marvellous palace which Constantine was building
for Sclerena, clouds began to gather over the life of the epicure. One
day, in the year 1044, as he rode with his guards at the head of a
religious procession, a cry broke from the crowd: “We don’t want
Sclerena as Empress, nor to see our lawful mistresses, Zoe and
Theodora, perish on her account.” The cry was a spark to the
spreading discontent, and the small troop of guards were
surrounded by a threatening mob. Fortunately for the Emperor, the
Empresses were watching the procession from the balcony, and they
sent troops to rescue him. Later, a discontented noble led some
Macedonian troops against the city, and encamped opposite the
Blachernæ gate. Constantine disdainfully ordered a chair to be
placed for him outside the gate, in order that he might see, and be
seen by, the rebels. For a time they were content to sing comic
songs about him—of which there must have been a good supply in
the city—then they made a dash and scattered his guards, and could
have penetrated into the city, possibly taken it, if they had not
foolishly retired. On such slender threads did crowns hang in that
singular Empire.
Sclerena relieved the growing discontent by a premature death,
apparently about the year 1045, and the superb palace which had
been intended for Constantine’s mistress was turned into a
monastery. Five years later Zoe closed her long and romantic career,
at the age of seventy. Constantine mourned for her as if she had
been a beloved child, and even pressed the Church to put her on the
list of the canonized; he may have read how St Theodora had won
the aureole largely by her freedom from jealousy. When it was
found, after a time, that some curious fungi had grown about her
monument, he insisted that they were heaven-sent assurances that
Zoe had been admitted at once into the company of the saints. The
Greek Church, however, was not persuaded to add Zoe to its quaint
list of the blessed, and few will reflect on the many events which
reveal her personality to us without admitting that, whether or no
she was guilty of the positive crimes attributed to her, she had little
or no moral feeling.
Constantine found consolation in the charms of a young Alan
princess who was detained as a hostage at Constantinople. The milk-
white skin and fine eyes of the unknown so fascinated him that he
gave her the imperial title and emptied the remainder of the treasury
upon her and the relatives who flocked to share her fortune. He was
by this time a miserable wreck of his former magnificent person, and
could not sit unaided on a horse, but the Court still rang with
laughter and buffoonery. His favourite, a man who had been raised
from the position of street buffoon to that of Court jester, became so
infatuated with his wealth and privileges that he dreamed of
possessing the pretty Alan princess and the purple. He was caught in
Constantine’s bedroom with a drawn sword. The Emperor asked why
he had attempted assassination, and, when the man said that he
had an irresistible passion to see himself in the crown and imperial
robes, burst into laughter and ordered the attendants to put them
on him. He returned to his position, and, to the amusement of
Constantine, made more open love than before to the fair Circassian
mistress. But the Emperor died in 1054, and his mistress returned to
her previous obscurity.
When it was seen that Constantine was failing, a number of the
nobles and officials conspired to put on the throne Nicephorus
Bryennius, but Theodora’s supporters forestalled the plot. They sent
a swift vessel for her and lodged her in the sacred palace before
their opponents could bring Bryennius from Bulgaria, which he
governed. She seems to have been forced out of affairs during the
later years of Constantine, and the sending of a boat implies,
apparently, that she had retired to the suburbs. She was still, in her
seventh decade of life, erect of form and clear in mind, and drastic
punishment was inflicted on the conspirators. She then began again
to control the affairs of the Empire as she had done in conjunction
with Zoe. She personally received ambassadors and heard trials, and
resumed her war on corrupt officials. Psellus is disdainful of her rule,
and unjust to her. The only grave defect we can recognize is that she
put the higher offices and commands at the disposal of men who
were less distinguished for ability than for devotion to her. A very
strong provincial aristocracy had by this time arisen in the Empire,
and from their vast estates a number of able nobles and officers
kept a discontented eye on the hierarchy of eunuchs at
Constantinople.
Theodora, conscious of her vigour, and sustained by the
prophetical assurance of a monk that she would wear the crown for
a long time, maintained her power for three further years, and then
became seriously ill. It is said that she chose an aged and feeble
noble of the city, Michael Stratioticus, to don the purple, but one is
rather disposed to see in the choice of Stratioticus the action of the
Court party, whose influence was threatened by the provincial
nobles. Theodora still confided in the monk’s prophecy; she had the
aged soldier brought to her sickbed and bound him by the direst
oaths to promise obedience to herself. She died a few days later,
however, on 30th August 1057, leaving the crown to the frail charge
of Michael VI. The historian must regret that Theodora had not a
larger opportunity to prove her value as a ruler and exhibit her
personality. She was a woman of great vigour and generally high
political ideals, and she incurs the reproach only of stooping at times
to the common Byzantine level in securing her power. It was not
she, but the contemptible Constantine, who emptied the treasury for
frivolous purposes, and, in spite of the light disdain of Psellus, her
rule compares most favourably with that of most of the Emperors.
CHAPTER XI

EUDOCIA

T
HE struggle which Theodora had foreseen was not long
deferred after her death, and Michael Stratioticus was
compelled, after a few months of feeble imperial
experiment, to retire to the private life from which he had been
unwisely drawn. The great territorial nobles—one might almost say,
the feudal nobles—concentrated upon the capital and put one of
their number, Isaac Comnenus, upon the throne. Isaac had in earlier
years married a Bulgarian princess, and her career as mistress of a
large provincial domain, and then as Empress of Constantinople,
suggests a very interesting study. Unfortunately, her husband’s reign
lasted only two years, and the events yield us only few and fleeting
glimpses of the new Empress.

Æcatherina, as the best contemporary authority, Nicephorus


Bryennius, calls her (though later writers often say Catherina),
descended from the Bulgarian royal family, which had fallen from its
high estate when “Basil the Bulgarian-slayer” had won a definitive
victory over the nation. Bryennius makes her a daughter of the King
Samuel, and we have in a later chronicle a picture of Samuel’s
daughters which would dispose us to imagine Æcatherina as a very
fiery and interesting personality. When, in the presence of Basil, they
were brought face to face with the woman whose husband had killed
their brother, the Emperor and his officers had great difficulty in
preventing a very violent and undignified scene. The dates, however,
make it improbable that Æcatherina was one of the daughters of
Samuel—others more probably suggest that she was his niece, or
grand-niece—and in character she seems rather to have been gentle
and religious. She was brought from her remote provincial home and
made Augusta, but she proved to be one of the quiet and retiring
Empresses who leave no mark in the chronicles. The only reference
to her is that, in 1059, she encouraged her husband, who had met
with a serious accident or illness, to resign, and she herself took the
veil of the nun. One suspects that her husband’s policy of curtailing
the funds of the luxurious and innumerable monks alarmed her, and
she was ready to believe that, as rumour maintained, the wild boar
which led him into grave peril in 1059 was no ordinary animal. He
resigned, and Æcatherina, changing her name to Helena, retired
with her daughter Maria to a quiet mansion, where they practised
monastic discipline and were esteemed so holy that Æcatherina was
eventually buried in the cemetery of the monks of Studion.
With the next Empress, Eudocia, we return to the more familiar
and more piquant type of Byzantine princess: the woman who unites
with her subservience to the Church a skill in casuistry which
protects her human inclinations from the harsher control of the
Church’s ascetic standards. Eudocia Macrembolitissa, or Eudocia the
daughter of Macrembolites, a distinguished noble of Constantinople,
had some beauty and no little wit, as well as good birth and
breeding. In the reign of Michael IV. and Zoe she had been wooed
and won by a handsome and learned, if not very warlike,
commander named Constantine Ducas, and had in the subsequent
twenty years of changing rulers borne three sons and three
daughters to her elderly husband. Constantine was at least ten years
older than she, and had no higher ambition than to be regarded as a
prince of letters and rhetoric. It must, therefore, have been an
agreeable surprise to Eudocia to learn, in 1059, that the retiring
Emperor had transferred his crown to her husband, and she was
henceforth to be the mistress of the sacred palace. She was then,
probably, in her later thirties. She was entitled Augusta, and the
imperial dignity was conferred also on her six children, of whom the
youngest was born after her coronation.
During the eight years of her husband’s reign Eudocia remained
a silent witness of his futility and unpopularity. He retained his
pedantry, and sought the laurels of learning and eloquence, while
formidable enemies threatened the Empire on every side. In 1067 he
perceived that his inglorious reign was about to end, and summoned
Eudocia, the nobles and the patriarch to his couch. The nobles were
commanded to swear to maintain the throne of Eudocia and her
sons, and Eudocia was compelled to swear a portentous oath that
she would not marry again. Possibly Constantine felt that he was not
imposing a very heavy sacrifice on a woman who approached her
fiftieth year, and it was plainly to the interest of his sons that she
should not marry. Eudocia signed the written oath, and it was
entrusted to the patriarch Xiphilin to keep in the great church.
The regency of Eudocia lasted about seven months, during
which she emulated the conduct of Zoe and Theodora. She received
ambassadors, heard trials and paid more direct and closer attention
to the affairs of the Empire than her late husband had done. Two
things, however, concerned her and illustrated the weakness of
woman-rule at Constantinople. The Turks and other hostile
neighbours were raiding the provinces with greater vigour, and the
nobles were making this a pretext for intrigue to replace Eudocia
with an Emperor. Before the year was out Eudocia decided to marry
again and sought a means of evading the oath which the patriarch
grimly guarded.
The story of her outwitting the patriarch is, as we find it in the
later chronicles, in the finest vein of Byzantine melodrama. She took
into her confidence one of the wiliest eunuchs of her Court, who
assured her that it was quite easy to induce the patriarch to release
her. This Xiphilin, the patriarch at the time, was himself as casuistic
as he was religious. Originally a noble, he had voluntarily embraced
the black robe of the monk, and had been withdrawn from the
monastery to rule the Eastern Church. He had in Constantinople a
brother named Bardas, whose gallantries and sybaritic ways were
notorious. When the eunuch proposed the subject of marriage,
Xiphilin sternly maintained that the oath was binding and that
Eudocia must remain a widow, but when the astute eunuch
regretted that such was his view, since it was his brother Bardas
whom Eudocia wished to marry, Xiphilin reconsidered the matter. It
is not for us to analyse his reasoning. It is enough that in a short
time he declared to the assembled Senators that the oath was
unjust and invalid, a mere wanton outrage on the part of a jealous
man, and he handed the precious document back to Eudocia to
destroy. His feelings may be imagined when, a few hours later, he
heard that the Empress was married, not to his brother, but to
Romanus Diogenes.
The contemporary writer Psellus gives a more sober version, but,
although Psellus was one of Eudocia’s chief ministers at the time,
there can be little doubt that his vanity and policy have somewhat
tempered the veracity of his narrative. Eudocia, he says, came to
him in tears to complain that the cares of Empire were an intolerable
burden for a single woman’s shoulders, and she wished to marry.
The story is, perhaps, not inconsistent with the story of her
outwitting the patriarch. In any case, the second marriage of
Eudocia had an element of romance.
In the state prison of Constantinople at the time was a
handsome young noble and commander named Romanus Diogenes,
who ran some risk of losing his head for high treason. Distinguished
by birth and in person, and a man of great spirit, he reflected that
the throne of the Eastern Empire had been reached by less able men
than he, and cherished a daydream of wearing the purple. At the
death of Constantine in 1067, when there was much discussion of
the empty throne and the imperial widow, he imprudently confessed
his ambition to those about him in the remote province of Thrace,
which he governed; he was denounced in the capital; and he was
brought in bonds to Constantinople and put on trial. He had then
completed his thirtieth year: a tall, comely, broad-shouldered man,
with the dark skin of a Cappadocian and very winning eyes.
Constantinople looked with sympathy on the manly, but impetuous,
young noble. He was connected by birth with the greatest families of
the Asiatic provinces, and he pleaded that it was only his concern for
the safety of the menaced Empire that had wrung from him words of
dissatisfaction. His treason was, however, apparent, and he was
found guilty and restored to jail.
Eudocia was probably present at the trial of Romanus, and noted
the handsome form and flashing eye. She professed afterwards that
the trial was unsatisfactory and must be revised, and the young
commander found himself acquitted and free to return to his native
province. The time was not yet ripe for the marriage project; in fact,
one of the historians states that Romanus was already married, and
went to join his wife and family in Cappadocia. About Christmas
(1067), however, he received an order from Eudocia to return to
Constantinople, and may or may not have been surprised to hear
that she proposed to marry and crown him. His wife and family
seem to have been deserted with great cheerfulness—unless we
24
prefer to regard the statement in the chronicle as an error —and
Eudocia secretly prepared for the marriage. Senators were bribed to
support the proposal, and, on 31st December, the patriarch was won
by the stratagem which I have already described. That very night
Romanus was introduced, fully armed, into the palace and secretly
wedded to the Empress, and on the first day of the new year the
young Emperor and his middle-aged Empress were ceremoniously
presented to the people. For a moment it seemed as if the fierce
Varangian guards were about to avenge what they regarded as a
violation of the oath to the dead Constantine, but Eudocia prevailed
on her elder sons to assure the guards that they had consented to
the marriage, and the trouble was averted for the time.
It was, however, in face of considerable hostility that Eudocia
and Romanus entered upon their task of governing the Empire. The
clergy were naturally hostile, since their leader had been tricked into
an ignominious concession; more distinguished nobles than
Romanus envied his elevation; and courtiers who were attached to
the fortunes of Eudocia’s elder sons regarded the new Emperor, and
the possible issue of the new marriage, with sullen distrust. Michael
Psellus, the historian who boasts that he guided Eudocia’s counsels
in regard to the marriage, is transparently hostile to Romanus, and
his historical work is largely responsible for the traditional prejudice
against that brave and spirited, but injudicious and unfortunate,
monarch. Psellus was not merely the chief student of philosophy in
Constantinople, but an ambitious and successful courtier. His great
repute in letters and philosophy gave him a commanding position in
25
the Court of Eudocia, who had herself some literary ambition, and
his secret and sinuous counsels must have deeply influenced the
later course of the careers of Romanus and Eudocia. A philosopher-
statesman was the great ideal which Plato, whose works he revived,
had urged upon the Greeks, but the fortunes of Psellus remain so
even throughout the various revolutions he outlived that one is
tempted to compare him rather with Talleyrand than with Plato’s
ideal.
EUDOCIA AND ROMANUS IV
FROM AN IVORY IN THE BIBLIOTHÈQUE NATIONALE, PARIS

Into this atmosphere of culture the robust Romanus was little


fitted to enter, and some disdain must have been felt of his
uncultivated ways. On the other hand, the brother of the late
Constantine, John Ducas, who bore the dignity of Cæsar and
jealously guarded the position of his nephews, was not less hostile
to Romanus. The boys had received the purple before the death of
their father, and the time was rapidly approaching when, with the
assistance of their uncle and Psellus, they might begin to exercise
their power. To this plan Romanus was a considerable obstacle.
When we further learn that Romanus was gravely conscious of his
duty to restore the strength and discipline of the army, and diverted
funds from the entertainment of idle citizens to the pay and
equipment of his troops, we realize that the life of the palace was
preparing for one more of those tragic revolutions which punctuate
the history of the Byzantine Empire.
From this Court atmosphere of pedantry and intrigue Romanus
turned to the field of battle; he would strengthen his position by
winning such laurels as his vigorous and warlike character seemed to
promise him. Two months after his coronation a fresh invasion of the
Turks was announced, and he led a large army out to meet them.
After nearly a year’s absence he returned with some report of
victories, but there had in the same year been heavy losses, and his
success was not decisive enough to override the intrigues of his
opponents. Already, we are told, he found Eudocia colder. Her
attitude is attributed to his arrogance and boastfulness; we may
suppose that it was just as much due to an instinctive irritation when
her robust husband strode into the philosophic atmosphere of the
palace with the smell of the camp clinging to him and the language
of war on his lips. In two or three months he was off once more to
the field, leaving Eudocia to her master of philosophy and her
brother-in-law. Into their hands she placed the more virile cares of
State, while she enlarged libraries, cultivated men of letters and
fostered the higher ambition of making verses. Her eldest son,
Michael, was associated with her in her cultural work.
When Romanus returned in the following winter, still without
decisive success, he seems to have concluded that it would be better
to remain in Constantinople, and the campaign of the third year was
entrusted to his generals, but in the spring of 1071 he again
prepared to take the field. Nothing but a crushing victory over the
enemies of the Empire would enable him to silence his enemies in
the Court and capital. Eudocia seems by this time to have wavered
between admiration of her young and manly spouse and repugnance
to his more robust standards of life. She was now certainly over fifty,
and had never been particularly sensuous, but we cannot doubt that
she had married Romanus for love and that that love was not yet
extinct. As he set out from port for his last crossing to Asia a
singular dark-plumaged pigeon circled his royal galley. He directed
that it should be caught and sent to the Empress; and it was said in
later years that Eudocia nervously recognized in the rare bird an
omen of the evil fortune that was about to befall her husband.
And in the course of the summer stragglers made their way
hastily to Constantinople with the news that Romanus had been
heavily defeated and his large army shattered. The Emperor himself
had been slain, some said, but at length there came men who had
seen him captured and borne away, a prisoner, by the Turks. The
hour of the malcontents had come, and a council was summoned to
discuss the situation. It was at once decided that no effort would be
made to save Romanus—some of the authorities declare that it was
the treachery of the Cæsar’s son, acting on the instructions of his
father, which led to the reverse—but the eldest son, Michael, should
be appointed ruling Emperor, together with his mother.
That Eudocia at once surrendered her husband becomes quite
clear from the subsequent course of events. The new administration
had hardly settled to its work when Eudocia received a joyful letter
from her husband announcing that he was free, and on his way to
Constantinople. How the Turk had entirely falsified his repute for
barbarity, treated Romanus as a brother king in misfortune, and
eventually released him on promise of a ransom, is a familiar and
attractive picture in the history of the time. Romanus was hastening
to the arms of his beloved wife. Eudocia is described by
contemporary writers as “distracted” and eager to consult those
about her as to her conduct. Of wifely feeling she did not exhibit one
sincere particle, and, however we may remind ourselves of the
inevitable coldness of a woman in her sixth decade of life, her
conduct is somewhat repellent. Had she known that the Cæsar was
bent on bringing her to a common ruin with her husband, she might
at least have purchased some loyalty to him, in the usual Byzantine
fashion; but she was either ignorant or powerless, and she accepted
the counsel that Romanus should be disowned and repelled by force
from his Empire.
John Ducas, however, concluded that the opportunity was
convenient for the removal of both Emperor and Empress. A decree
was issued to the provinces to arrest the advance of Romanus, and
the guards were marshalled. At this date the mercenary troops in
charge of the palace were the famous and formidable Varangian
guards, in whom modern authorities recognize the blue-eyed giants
of distant Scandinavia and even of Britain. Romanus had favoured
the native troops of the Empire rather than these foreign
mercenaries, and they at once accepted the command of the Cæsar.
One half of them went to the apartments of Michael, and declared
him sole Emperor of the Romans; the other body went in search of
Eudocia, with orders to transfer her to a monastery.
Eudocia at once concluded that the end of her rule had come
when she heard the jubilant clash of axe on shield, the deep guttural
voices, raised in song, of the northern soldiers, and their heavy tread
across the gardens and terraces. Fearing for her life, she hid herself
in some sort of hut in the grounds of her palace, but the door was
presently flung open and she looked on the fierce hairy faces and
shining weapons of the Varangians. She was prostrate with terror
when the Cæsar arrived, to give her the comparative consolation
that her life would be spared, but her empire was over. From the
palace, spoiled of all the ensigns of royalty, we follow her along the
short and painful route that we have seen so many proud rulers of
the sacred palace take. At the Bucoleon quays a swift galley waited
to take her to the Asiatic shore, where she was lodged in a
monastery which she herself had founded. A further message soon
came, ordering her to take the black veil, and the frail and
unfortunate woman bade farewell to all the glories of imperial life. It
was only four years since she had been left in control of the Empire
by her first husband.
Shortly afterwards she was summoned to bury Romanus, and
with him the last flickering hope of a return to power. He had
collected an army and resolved to fight for his throne, and the
troops of Ducas at length pinned him in a town of Cilicia. In order to
end the civil war John now sent an assurance that the life of
Romanus would be spared if he would resign his claim and enter a
monastery; nay, three archbishops were sent to give him a solemn
testimony that John had sworn and would fulfil his oath. Frail as the
most formidable oaths had become in Eastern Christendom,
Romanus opened the gates and yielded to the sons of the Cæsar.
The rest of the story is a chapter of nauseous horror, and concerns
us, fortunately, only in outline. Romanus was conveyed across Asia
Minor, in the robe of a monk, with studied insult. Most of the
chroniclers affirm that poison was administered to him, but that his
powerful constitution prevented it from doing more than add to his
misery. At length his eyes were cut out with more than ordinary
brutality, the roughest and most elementary attention to his bleeding
sockets was refused, and he was borne once more on a mule, dying
by inches in the most ghastly conceivable fashion, across Asia Minor.
He reached the island of Prote in time to die on the soil that was
already watered by so many imperial tears, and the chroniclers add
that Eudocia gave a splendid funeral to the remains of the man
whom she had transferred from the jail to the palace, less than four
years before, in the full pride of a magnificent manhood.
I have said that with the remains of Romanus she buried her last
hope of returning to power, yet some seven years afterwards a
strange message reached her in her cloister, recalling the memory, if
not the hope, of imperial power. Her son Michael proved an
ineffective ruler. The tradition of culture which had lingered in the
palace since the days of Psellus absorbed all his energy, and he
could not be diverted from the dialogues of Plato or the iridescent
dreams of Plotinus by mere conspiracies against his throne or
invasions of his Empire. Indeed, it was with difficulty, sometimes,
that they could drag him to table or persuade him to refrain from
spending the night over his books. The irony of the situation was
that, while the Greek writings over which he lingered urged that a
profound study of philosophy was the fittest education of monarchs,
Michael remained as helpless and heedless as a boy, precisely on
account of his studies. Fortunately, he had the casual inspiration to
call to the palace a wily eunuch, named Nicephorus, who become
the virtual ruler. Nicephoritzes—as the people, using the diminutive
form of his name, called the pale and shrunken little eunuch—soon
displaced the Cæsar John, and, as was the invariable custom of his
kind, enriched himself at the expense of the impoverished and
decaying provinces.
Under Nicephoritzes Eudocia had no chance of a return to power.
He had endeavoured to persuade her first husband, the Emperor
Constantine, that she was unfaithful to him, and had been driven
from office during her regency. But the Empress’s quarters in the
palace were not vacant; a new type of Empress was added to the
long and varied gallery. Shortly before his accession to the supreme
throne Michael had married a princess of one of the tribes that had
settled in Asia Minor. The father of the Empress Maria is conflictingly
described as a king of the Iberians and the Alans, and is said to have
been a ruler of great fame and power; but he is not named, and it
seems that he was not powerful enough to avert or temper the
tragedy of his daughter’s career. Her dowry had been her beauty. I
have complained at times of the lamentable indifference of the male
historians of Constantinople to the physical features of the
Empresses, and the lack of portraits which might bring the living
figure with any fulness or accuracy before the imagination. We now,
however, approach a period, the history of which has been written
for us by a woman, the famous Anna Comnena, and her pen happily
wanders at times back to the age of Eudocia, of which her husband,
Nicephorus Bryennius, was the chief historian.
Unhappily, the art of which Anna Comnena was so patently
proud did not include skill in portraiture. Maria was the most
beautiful woman of her time, and, although her interests become
opposed to those of Anna and her family, and the learned princess
was capable of malignant hatred, Anna Comnena rises to the height
of superlative when her pen delineates the figure of Maria. Her grace
of form and beauty of face were beyond the artist’s power to
convey; though one must add that Anna not infrequently uses that
formula, in order to enhance the artistic wonder of her own
descriptions. Maria, she says, was tall and graceful as a cypress; her
body was white as snow, save for the roses that bloomed in her
cheeks, and the luminous blue eyes which shone beneath the
perfect and lofty arch of her auburn eyebrows. To this vague poetical
description we may add at once that the beautiful young princess
was not wholly devoid of the spirit of her tribe, and was prepared for
romantic adventure in support of the imperial dignity.
The seven years of Michael’s reign do not interest us. The
Emperor lived in the remote solitude of his exalted studies; Maria
enjoyed the superb luxury of her position, and brought a prince into
the world for the greater security of her throne; Eudocia languished
in the royal monastery of the Virgin across the straits. Usurpers rose
and fell, and the defrauded people spoke with bitterness of the
young pedant who let his ministers rob them while he studied the
divine maxims of Plato. Another princess, daughter of Robert of
Lombardy, was introduced from the West, but she was, like Maria’s
son, to whom she was betrothed, a child of tender years, looking
with strange blue eyes on the vast palaces she would one day
govern—they said—and the boy who shyly shrank from her
companionship.
At last, in 1078, a more fortunate rebel advanced on
Constantinople, the clergy and nobles were bribed to espouse his
cause, and Michael fled to the Blachernæ palace in the suburbs.
Maria accompanied him, and what we know of her character
emboldens us to fancy her urging the distracted scholar to draw a
sword on behalf of his throne. His friends, however, found it
impossible to move him, and, yielding to the usurper, he was
conducted on an ass to the monastery at Studion, where he might
prosecute his studies with even greater leisure. The new Emperor
had so genial a disdain for him that he made him titular Bishop of
Ephesus, and allowed him to return and live in the capital.
Maria, in accordance with custom, entered the suburban
monastery at Petrion. She did not, however, take the vows of the
religious life, and it was not long before the interesting news came
that the new Emperor designed to marry her. Nicephorus
Botaneiates was an elderly voluptuary, who had seized the throne
only because so little energy was needed for the task. For the
administration of public business he had two slaves of his own
household, of Slavonian extraction, who at once put an end to the
life of Nicephoritzes and diverted the stream of gold to their own
pockets. For their master the pleasures of the table and the couch
sufficed. He had brought to the throne an obscure Empress named
Berdena, but she died shortly afterwards, and the aged Sybarite
consulted his ministers. To their cold and impartial judgment it
seemed that political considerations must rule the choice and they
were divided between the claims of Maria and those of Eudocia. It is
true that Nicephorus had been twice married, that Eudocia was a
nun, and that Maria was not yet a widow; but such difficulties were
never beyond the casuistic resources of the Constantinopolitan
clergy. The Emperor must marry, since the sacred ritual of the Court
demanded the presence of an Empress.
The politicians favoured the suit of Eudocia, and she was
actually informed that Nicephorus wished to marry her, and
expressed her cordial willingness to sacrifice her monastic estate in
view of such august considerations. Nicephorus, however, was, as I
said, a Sybarite, and even advanced age did not blur his experienced
eye to the charms of Maria. We may, therefore, suppose that
Nicephorus was neither surprised nor pained when a certain very
holy monk appeared at the monastery of the Virgin and sternly
forbade Eudocia to quit her black robe. It may be that the monk was
one of the chaplains of the monastery; it is at least clear that his
zeal did not take him to the monastery at Petrion, where Maria
resided. The beautiful young Empress was recalled from her prayers
and fasts and conducted to the side of the Emperor in the palace
chapel. The patriarch, who seems to have had some scruples, was
not summoned to perform the ceremony, and Nicephorus noticed
with irritation that the priest who was called hesitated to come to
the sanctuary; Nicephorus had no dispensation for a third marriage,
and Maria’s husband still lived. A courtier, however, had foreseen the
difficulty and had a more accommodating priest at hand. The
irregular knot was tied, or regarded as tied, and Maria returned to
enjoy, with her son, the pleasures of the Emperor’s luxurious Court.
It is, perhaps, no alleviation of the conduct of Maria, in
purchasing her crown by an invalid marriage to an elderly sensualist,
to say that—the chroniclers assure us—quite a number of noble
ladies at Constantinople were eager to be chosen. Eudocia, her
youngest daughter, Zoe, and many other ladies had been pressed
upon the notice of Nicephorus. It is merely one more indication of
the inferiority of character, both in men and women, in the Byzantine
Empire. But Maria was not destined to enjoy long the throne which
she had purchased. Contemptible as the reign of Michael had been,
it was succeeded by one far more contemptible, and sullen murmurs
filled the palace and the city. Men told each other how the aged
Emperor, who ought to be thinking of eternity, changed his splendid
robes ten times a day, anointed his jaded frame with the most costly
unguents, and sat down, day after day, to the most superb banquets
that the Empire could afford; while the two barbaric slaves whom he
had made his chief ministers ground the despairing provinces and
disgusted the nobles. Within a year or two of Maria’s return to
power, the customary, inevitable revolt arose, and she was driven
back to her monastery.
This revolution, however, introduces us to the strong women of
the Comnenian house and must commence a fresh chapter. Of
Eudocia we hear no more. If we accept the statement of one of the
chroniclers, that she had married in the reign of Michael IV. (1034–
1041), she must now have reached her seventh decade of life, and
would probably not long survive her last disappointment. Her
readiness, in her later sixties, and after seven years of monastic life,
to accept the embraces of a roué like Nicephorus, in return for the
crown, is a sufficient measure of her character; her violation of her
oath to her first husband, and her desertion of her second husband,
point to the same feebly vicious and unattractive type of personality.
Through the favour of Nicephorus she was permitted to leave the
suburban monastery, and spend her last years in considerable
comfort in the city.
CHAPTER XII

IRENE AND ANNA COMNENA

T
HE distinguished family of the Comneni has already made its
appearance in our narrative. It may be recalled that the last
chapter opened with a march of the great provincial nobles
upon the capital, and the placing of one of their ablest
representatives, Isaac Comnenus, upon the throne. Isaac’s brave life
had ended in heroic foolishness. Terrified by an apparition, he
embraced the monastic life, ignored the natural desire of his brother
John to succeed him, and handed the crown to the Ducas family.
During the reign of Eudocia the widow of John Comnenus, Anna,
remained in Constantinople to guard the fortunes of her children and
eventually to help them to secure the throne. She was a woman of
the old Roman build, rather than Byzantine; strong, ambitious, able
and despotic. The Cæsar John Ducas looked on her with just
suspicion, and accused her of treasonable correspondence with
Romanus, when he was struggling to regain his throne. She boldly
asserted that the letters were forged, and brandished an image of
Christ in the eyes of her judges; but it was expedient to condemn
her, and she passed to the melancholy Princes’ Islands.

Michael the Scholar released her as soon as Diogenes was dead,


and she returned to Constantinople, to watch and work. She had
something of the spirit of her father, who had sent so many of the
enemy to the land of shades that he had won the name of Alexius
Charon: her mother had been of the great family of the Delasseni.
The feebleness of Michael and the insipidity of Nicephorus gave
promise of a successful revolution, and Anna and her two sons were
shrewd enough not to force the opportunity. The youth had first to
learn the mastery of legions and to marry. There were, in fact, four
women in Constantinople, all able and ambitious, who sought the
throne for their children, and a stupendous amount of intrigue must
have been expended. The four were: Anna Comnena, the Empresses
Eudocia and Maria, and the wife of Andronicus, son of the Cæsar
John Ducas. Andronicus had been fatally wounded in war, and
condemned to a lingering death, and his wife pressed the Cæsar to
find good alliances for her three daughters. She was one of those
virile and beautiful Bulgarian princesses who had found the way to
Constantinople, and her eldest daughter, Irene, was now just
marriageable.
The wife of Andronicus—we do not know her name—shrewdly
concluded that an alliance with the Comneni would best serve her
ambition, and she pressed her father-in-law to bring about a
marriage between Irene and Alexis, the elder of Anna’s two sons.
Alexis was a very promising and successful commander who had
recently lost his first wife, and he was not unwilling to wed the fair
Irene. Anna Comnena (the younger) describes the pair for us, with
her usual verbosity and inexactness, premising that it is beyond the
power of art to reproduce their comeliness. Alexis was, it seems, a
man of medium height, with very broad shoulders and massive
chest, eyes of “terrible splendour,” and a look that was “at once both
truculent and bland.” He seems, in fact, to have been a very ordinary
young man, with an extraordinary capacity for ruse and intrigue.
Irene (Anna’s mother) was, of course, a paragon. Her face was “like
the moon,” though not quite so round, and her rosy cheeks and fine
blue eyes make the simile somewhat weak; her look, like that of her
husband, was “at once sweet and terrible”—the look of “a Minerva of
heavenly splendour”—and calm and storm succeeded each other, as
on the sea, in her expressive blue eyes; her arms and hands were
like carven ivory, and her constant gestures extremely graceful. In
other words, Irene was a very pretty maiden of thirteen summers at
the time, with a large share of the spirit and temper of her Bulgarian
mother. These fragments of Anna Comnena’s art may serve to
illustrate Gibbon’s indulgent complaint that it is more feminine than
the artist herself.
The prospect of so significant a marriage released a fresh flood
of intrigue. Anna, the mother of Alexis, remembered that it was John
Ducas who had driven her into exile, and would not hear of a match
with his daughter-in-law. The Emperor Michael regarded the
marriage with distrust; his brother Constantine wanted to marry
Alexis to his sister Zoe, Eudocia’s youngest daughter. Through this
thicket of obstacles and intrigues the wife of Andronicus fought her
way with spirit, and not a little bribery, and the marriage took place.
We may assume that this was in the second or third year of
Nicephorus, when Irene, who was only fifteen at her coronation,
cannot have been more than thirteen or fourteen years old.
The Empress Eudocia had now played her last card, and
resigned herself to the life of the monastery; it remained to secure
the favour of the lovely Empress Maria. Isaac Comnenus had married
her cousin Irene, and had therefore the entrée of her palace. The
Slavonian ministers of Nicephorus watched him and his brother with
concern, but he won the affection of Maria and, by generous
distribution of money, the service of her eunuchs. It was presently
announced that the Empress Maria proposed to adopt the successful
young commander of the troops, Alexis Comnenus, and when this
ceremony had been performed both brothers were at liberty to make
lengthy visits to the Empress. It is not difficult to accept the rumour
that the relation of Alexis to his “mother” was not entirely filial.
Alexis was no ascetic, and he notoriously strayed from his girl-wife.
On the other hand, Maria had not shown much delicacy in marrying
the white-haired sensualist, and the privilege of intimacy with a
handsome young general of thirty-seven, her eunuchs being bribed
in his and her favour, would be appreciated by her. Her mind was not
strong and penetrating enough to see through the trickery of Alexis.
He posed as an unambitious general, loyally devoted to her reign
and that of her son.
The Emperor Nicephorus probably felt that the young men would
await the natural termination of his imperial orgies before seizing the
throne, and seems to have regarded them with a certain genial
indifference. His ministers, however, knew that their fortunes were
ruined if Alexis came to the throne, and they insisted that
Nicephorus must name a successor. He chose his nephew, a
handsome young noble named Synadenus. Maria was now seriously
alarmed, since the accession of Synadenus would mean the
monastery for her and, possibly, death for her son, and she allowed
the Comneni to witness her tears. They were, they said, devoted to
her cause. Nay, they swore on the holy cross that they would
acknowledge no rulers but Maria and her son, and she promised, in
return, that they should be informed of any step that might be
contemplated against them in the palace. I am following, almost
entirely, the narrative of Anna Comnena, who enlarges with the most
candid pleasure on the deceit of her father, and assures us that her
grandmother, Anna, was the soul of the plot. In the palace of the
Comneni councils were held daily, and the virile mother directed the
movements of her sons. It was a time of great anxiety. One night
Nicephorus invited Alexis and Isaac to his banquet, and Anna depicts
them nervously glancing round them during the meal for the guards
or assassins who might have been summoned to despatch them. But
Alexis, a master of ruse and insinuation, won the Emperor, and,
when a charge of treason was afterwards brought against him, he
easily cleared himself.
At last a message came to the mansion of the Comneni from
Maria that Barilas (one of the Slav ministers) intended to seize the
throne and put out the eyes of Alexis; and it was decided that the
time had come for action. Alexis hastily made a tour of the city,
persuading some, bribing others, until he had a large number of
officers and Senators bound by secret oath to support him. Anna
meantime made preparations for the flight of the family during the
night. The chief weakness of their position was that a young relative
of the Emperor had recently married a young girl of their family, and
lived, with a tutor, in an outlying part of their mansion. Anna,
regarding the tutor as a spy, locked them in their rooms when they
were asleep, and before dawn the whole Comneni family set out on
foot to cross the city. At that hour of the night there was little watch
in Constantinople, and the nervous band—the mother, the two
brothers with their wives, children, and sisters, and a few servants—
passed safely and silently down the colonnaded main street as far as
the Forum of Constantine, where horses awaited the men. They
bade each other farewell in the darkness of the early spring
morning, and the brothers galloped to the Blachernæ palace, where
they broke into the stables, chose the swiftest horses, hamstrung
the rest of the horses, and fled to the army which awaited them in
Thrace.
The women and children made their way noiselessly back along
the Mese to the cathedral. As they went along the street, the glare
of a torch appeared in the distance and they found themselves
inconveniently accosted by the tutor spy. Anna kept her presence of
mind, however. They had heard, she said, that they were accused of
some crime and they were going at once to St Sophia, but as soon
as the day broke they would go to the palace to demand justice, and
she begged the tutor to go on to the palace to announce their
intention. As soon as he had gone, they made for the house of
Bishop Nicholas, an annexe of the cathedral into which fugitives
were admitted during the night. Rousing the doorkeeper, they
announced themselves—they were all heavily veiled—as a party of
women who had just landed at the quays from the east, and who
would render thanks to the Almighty before repairing to their homes.
They were admitted to the church, and, when the officers of the
infuriated Emperor arrived, in the early morning, they found that
nothing less than a violation of the sanctuary would put the women
in the power of Nicephorus. Anna, in fact, clung to the gates of the
sanctuary, and exclaimed that the soldiers would have to cut off her
hands to remove her from the church, as the Slav ministers
threatened. Isaac’s wife Irene, an Iberian princess like her cousin
Maria, followed the example of her mother-in-law, and we must
imagine the younger Irene and the children standing by, with large
and tearful blue eyes, taking their first lesson in Byzantine politics.
Nicephorus temporized, and swore to spare their lives. Anna
shrewdly stipulated that his oath should be taken on the large cross
which the Sybarite Emperor always wore, and, when this had been
brought and the oath guaranteed to them, the women passed from
the church to the palace-fortress-monastery at Petrion, on the
Golden Horn. There they were soon joined by the wife and mother-
in-law of George Paleologus, a dashing young commander who had
fled with the Comneni, and, by sharing their delicate meats and
wines liberally with their jailers, they secured a constant account of
the progress of the insurgent brothers.
They heard presently that Alexis and Isaac had safely reached
the camp in Thrace, and that it had needed only a little further
intrigue on the part of Alexis for the troops to proclaim him Emperor.
The next news of importance was that the brothers were encamped
with their troops on the higher ground without the city walls, and
Nicephorus was distracted and terrified. But we may tell in few
words the success of the Comneni. The formidable walls of
Constantinople were held by the Varangian guards and Immortals,
on whose blind fidelity a ruling (and paying) Emperor could always
rely. But the extravagance of Nicephorus had in three years
exhausted the treasury—its doors stood open for any man to enter
the empty building—the troops were few, and uncertain mercenaries
had to be enlisted in the defence. Alexis bribed the German soldiers
who held the tower overlooking the Blachernæ gate, and at dawn of
Maundy Thursday (1081) his troops poured into the city.
It is one of the few points in favour of Alexis that he here made
a very human blunder which might have cost him his life and his
ambition. Instead of holding his troops to scatter the guards, who
had retreated upon the palace, he rode at once to Petrion to see
that the women were safe, and his soldiers—a motley and savage
crowd of Thracian and Macedonian mercenaries—spread with
fiendish delight over the city, violating nuns in the monasteries and
burdening themselves with wine and loot. Paleologus saved them by
a bold and crafty seizure of the fleet, cutting off the Emperor’s

You might also like