Nature Inspired Optimization Techniques for

Image Processing Applications Jude Hemanth pdf

download

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

Download more ebook from https://textbookfull.com

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

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/

Human Behaviour Analysis Using Intelligent Systems D.

Jude Hemanth

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

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/

Image Processing and Communications Techniques

Algorithms and Applications Micha■ Chora■

https://textbookfull.com/product/image-processing-and-
communications-techniques-algorithms-and-applications-michal-
choras/
Nature-Inspired Algorithms and Applied Optimization 1st
Edition Xin-She Yang (Eds.)

https://textbookfull.com/product/nature-inspired-algorithms-and-
applied-optimization-1st-edition-xin-she-yang-eds/

Nature Art Workshop Tips Techniques and Step by Step

Projects for Creating Nature Inspired Art 1st Edition
Sarah Lorraine Edwards

https://textbookfull.com/product/nature-art-workshop-tips-
techniques-and-step-by-step-projects-for-creating-nature-
inspired-art-1st-edition-sarah-lorraine-edwards/

Nature-Inspired Optimizers: Theories, Literature

Reviews and Applications Seyedali Mirjalili

https://textbookfull.com/product/nature-inspired-optimizers-
theories-literature-reviews-and-applications-seyedali-mirjalili/

International Conference on Intelligent Data

Communication Technologies and Internet of Things ICICI
2018 Jude Hemanth

https://textbookfull.com/product/international-conference-on-
intelligent-data-communication-technologies-and-internet-of-
things-icici-2018-jude-hemanth/

Signal and Image Processing Techniques for the

Development of Intelligent Healthcare Systems E. Priya

https://textbookfull.com/product/signal-and-image-processing-
techniques-for-the-development-of-intelligent-healthcare-systems-
e-priya/
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: kacprzyk@ibspan.waw.pl

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: jainlakhmi@gmail.com; jainlc2002@yahoo.co.uk
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
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: appu123kumar@gmail.com
A. Lenin Fred
H. Ajay Kumar
Mar Ephraem College of Engineering and Technology, Elavuvilai, India
e-mail: leninfred.a@gmail.com
H. Ajay Kumar
e-mail: ajayhakkumar@gmail.com
P. Sebastin Varghese
Metro Scans and Research Laboratory, Trivandrum, India
e-mail: sebastin464@gmail.com

© 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
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
S¼
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
1 Firefly Optimization Based Improved Fuzzy Clustering … 11

trapping at local minima. The improved FCM based on firefly optimization replaces
the classical techniques of random initialization.
Prior to filtering, the medical image film artifacts are eliminated by a statistical
technique coupled with convex hull computational geometry [33]. The threshold
value determined by standard deviation technique was used for the binarization of
input image. The binarized image was then subjected to connected component
labeling for the elimination of patient details and technical information. The convex
hull of the resultant image was multiplied with the original image for the generation
of artifacts removed image. The preprocessing of input image was performed by
Non-Linear Tensor Diffusion (NLTD) filter prior to segmentation [34]. The NLTD
ensures good edge preservation since the smoothing is heterogeneous and
non-noisy pixels are not disturbed.
The computation complexity was minimized by reducing the pixel count for the
processing by segmentation algorithm.

Rp ¼ randperm ðLÞ

The parameter Sp represents the pixels taken for optimization, here in this work
50% of the total pixels are taken. The L represents the total pixel count of the image
to be segmented and randperm function returns a row vector depicting a random
permutation of the integers from 1 to n.

Sn ¼ ceil L  Sp

The Sn represents the number of pixels selected for optimization and the function
below represents the subset of pixels chosen for optimization process

X2 ¼ X ðRpð1 : Sn Þ; :Þ

The optimization of the objective function relies on the brightness and move-
ments of the firefly. The firefly algorithm starts by initializing the population of
fireflies. The intensity of light emitted by the firefly estimates the movement of the
fireflies. The algorithm works in the iterative fashion. The intensity of ith firefly is
compared with the jth firefly as follows
if bðiÞ [ bðjÞ
firefly j move towards firefly i
else
firefly i move towards firefly j
12 S. N. Kumar et al.

The incorporation of firefly algorithm has significantly improved the segmen-

tation results. There were four stages of Improved FCM-Firefly segmentation
algorithm.
i. Initialization phase
ii. Intensity calculation phase
iii. Movement calculation phase
iv. FCM algorithm phase.
The goal of incorporating firefly optimization in FCM is to minimize the
objective function with a global minima value. The cluster centers represent
the decision parameters to minimize the objective function. The initialization of the
firefly population is as follows

yif ¼ yi1 ; yi2 ; . . .; yij ; . . .yid 2\j\C

Each firefly in the population is represented by using the above equation. Where
yij represents the jth cluster centre.
The population of the fireflies are initialized and randomly distributed in search
space. The position of firefly depicts the possible solution (centroids) for the
clustering problem. In this phase, the parameter like b0 ; c; a and maximization
iteration are also initialized. Once the initialization process is over, the intensity of
each firefly is determined by estimating the distance between the position of the
firefly and the entire data in the dataset. The minimum distance value among the
population with respect to data from the dataset is considered. The intensity value of
each firefly is determined based on the sum of minimum distance with respect to the
data from the dataset.
The expression for determination of intensity is as follows

X
n
bðFFj Þ ¼ di
i¼1

where FF represents firefly, di represents the minimum distance value for a par-
ticular firefly.
The brightness of the fireflies indicates the movement of the fireflies in the search
space. The intensity of fireflies is compared to determine the new position. The
difference in the brightness triggers the movement. The firefly optimization is
employed in the FCM algorithm to enhance the clustering operation. The new
position of the entire swarm of the fireflies is determined by the FCM operator
based on the current intensity value.
The FCM-Firefly algorithm is carried out through the updation of the mem-
bership value uij and position of the firefly yj using the below equations
1 Firefly Optimization Based Improved Fuzzy Clustering … 13

1
Uij ¼  2=f 1 ; 1iN
Ps kyi fj k
k¼1 kyi fk k

where Uij depicts the degree of membership of yi in the firefly j, degree of fuzziness
f = 2 and yi is the data associated with the firefly under study.
PN m
i¼1 uij
xi
F j ¼ PN m
i¼1 uij

The Fj represents the solution after applying FCM in the firefly j.

The new position of Firefly is determined and the intensity value is updated. The
fixed number of iterations will be provided and at the end of the iteration, the best
solution was determined.

1.3 Results and Discussion

The algorithms are developed in Matlab 2010a and tested on CT abdomen data sets.
The system specifications are as follows; Intel Core i3 processor of 3.30 GHz with
4 GB RAM. In the scenario of medical image segmentation, fixing the number of
clusters is cumbersome, since it cannot be initialized roughly by viewing the image.
The validation metrics are employed for the optimal cluster selection.
This is performed in 3 steps
i. The parameters of clustering algorithm except the cluster number is fixed.
ii. The cluster number is varied from an initial value of 2 to an upper limit (max).
The data partition is carried out for each cluster number.
iii. The cluster validity indexes are applied on the data partition obtained from the
previous stage for evaluation. Based on the values of CVI’s, the cluster
number selection is done.
The terminologies used in the formulation of cluster validity function are as
follows
N: the count of data objects for clustering
f: the fuzzifier factor that represents the level of cluster fuzziness
u: the ith data object, 1  i  N
P: the number of clusters
Cp : the pth cluster, 1  p  P
 
Cp : the count of data objects in the p-th cluster
Vp : the centroid of the p-th cluster
ku  vk: the distance between a pair of data objects
lip : the membership degree of ui corresponding to Cp .
14 S. N. Kumar et al.

The FCM algorithm is an iterative technique in which the pixels are grouped into
a cluster based on the membership degree through the minimization of the objective
function.

M X
X P  2
lipf ui  Vp  ; f 1
i¼1 p¼1

The number of the cluster is taken initially P and randomly P centroids are
selected. The objective function represented above is optimized in an iterative
fashion by the updation of lip and Vp as follows

1
lip ¼  2
PP kui vp k f 1
2

kuj vc k
j¼1 2

PN f
i¼1 lip ui
Vp ¼ PN f
i¼1 lip

The iteration terminates when, kUT þ 1  UT k\ 2, where UT ¼ lip represents

the matrix comprising of all lik 0 s. T is the number of iteration and 2 is a threshold
specified by the user. The clustering validity metrics are used to estimate the quality
of clustering result. The partition coefficient (PC) and partition entropy (PE) is
based only on the membership values of fuzzy partition dataset. The criteria for
optimum cluster number selection is the maximization of (PC) or minimum of PE.
The issues in the performance metrics, PC or PE is that they do not consider the
geometrical properties of the dataset.
Xie Beni’s index (XBI) and Fukuyama’s and Sugeno’s index (FSI) are also
widely used classical CVI’s. XBI and FSI focus on the characteristics, compactness,
and separation. The numerical part of the expression XBI in Table 1.1 represents
the compactness of fuzzy partition, the denominator part represents the strength of
separation between the cluster for optimal clustering. The value of XBI should be
minimized for optimum cluster number selection. The expression for FSI in
Table 1.1 comprises of two terms. The first term represents the compactness
measure and the second term represents separation measure. Though FSI and XBI
consider the inter-cluster information, geometrical properties are not considered.
The DB index was obtained by the mean of cluster similarities. For each cluster
P, the similarity between P and all other clusters are determined. The term Sp is
represented as follows

1 X  
Ui  Vp 2
Sp ¼
jCP j Ui 2Cp
Other documents randomly have
different content
 CHAPTER II
"And I shall not see you for a whole month, my precious pet!" Lord
Algy whispered, as the train was approaching Charing Cross, at
about eleven o'clock on the Monday night of the return journey. "I
don't know how I shall bear it, but you will write every day, won't
you?—Promise me, darling——I wish now that I had not taken first
leave and arranged to shoot with my brother-in-law next week."
His arm still encircled her, and her ashen-hued head leaned against
his shoulder, so that he could not see the expression in her sombre
eyes. It was that of an animal in pain.
"No, I shall not write, Algy, and you must not, either—we have had a
divine time, and I shall never forget it. But it is stupid to write—what
good would it be to either of us?"
He pleaded that he would not be able to live without a word—after
the three days of perfect bliss they had enjoyed—and, of course,
they would enjoy many more, when he returned from Wales—!
Katherine Bush did not argue with him—of what use since her own
mind was entirely made up? She just let him kiss her as much as he
desired without speaking a word, and then she arranged her hat and
veil, and was demurely ready to get out when the train should draw
up at the platform.
Lord Algy could not have been more loverlike. He was really feeling
full of emotion and awfully sorry to part. She had been so wonderful,
he told himself. She had enjoyed the whole thing so simply, and was
such a delightful companion. She had not asked any silly questions
or plagued him with sentimental forever-and-ever kinds of
suggestions, as lots of girls might have done with her limited
experience of these transitory affairs. She had accepted the situation
as frankly as a savage who had never heard that there could be any
more binding unions. He really did not know how he was going to
stand a whole month of separation, but perhaps it was just as well,
as he was on the verge of being ridiculously in love, and to plunge
in, he knew, would be a hopeless mistake. She was a thousand
times nicer and more interesting than any girl he had ever met in his
life. If she had only been a lady, and there would not be any row
about it, he could imagine any fellow being glad to marry her.
She was not at all cold either—indeed, far from it—and seemed
instinctively to understand the most enchanting passion—He thought
of Mademoiselle de Maupin again—and felt he had been as equally
blessed as D'Albert. She would make the sweetest friend for months
and months, and he would rush back from Wales the moment he
could break from his family, and seek solace in her arms—he would
have got himself in hand again by then, so as not to do anything
stupid. He always meant to be very, very good to her, though. Thus
he dreamed, and grew more demonstrative, clasping her once again
in a fond farewell embrace, during the last available moment, and
his charming blue eyes, with their brown curly lashes, looked half full
of tears.
"Say you love me, darling," he commanded, wishing, like all lovers,
to hear the spoken words.
Katherine Bush was very pale, and there was concentrated feeling in
her face which startled him. Then she answered, her voice deeper
than usual:
"Yes—I love you, Algy—perhaps you will never know how much. I do
not suppose I will ever really love anyone else in the same way in
my life."
Then the train drew up at the station.
The people all looked unreal in the foggy October air under the
glaring lights—and the whole thing appeared as a dream indeed
when, half an hour later, Katherine sped through the suburban roads
to Bindon's Green, alone in the taxi. Lord Algy had put her in and
paid the man liberally, and with many last love words had bidden her
good-night and—au revoir!
So this chapter was finished—she realised that. And it had been
really worth while. An outlook had opened for her into a whole new
world—where realities lived—where new beings moved, where new
standpoints could be reached. She saw that her former life had been
swept from her—and now, to look back upon, appeared an
impossible tedium. She had mastered all the shades of what three
days of most intimate companionship with a gentleman could mean,
and the memory contained no flaw. Algy's chivalry and courtesy had
never faltered; she might have been a princess or his bride, from the
homage he had paid her. Dear, much-loved Algy! Her passion for him
was tinged with almost a mother love—there was something so
tender and open-hearted about him. But now she must take stern
hold of herself, and must have pluck enough to profit by what she
had learned of life—Though to-night she was too tired to do more
than retrospect.
Oh! the wonder of it all!—the wonder of love, and the wonder of
emotion! She clenched her cold hands round the handle of her little
valise. She was trembling. She had insisted upon his keeping the fur-
lined coat for the present. How could she account for it to her family,
she had argued? But she never meant to take it again.
No one was awake at Laburnum Villa when she opened the door
with her latchkey, and she crept up to her little icy chamber under
the roof, numb in mind and body and soul—and was soon shivering
between the cotton sheets.
Oh! the contrast to the warm, flower-scented bedroom at the
Palatial! And once she had not known the difference between linen
and cotton!
She said this over to herself while she felt the nap—and then the
tears gathered in her eyes one by one, and she sobbed
uncontrollably for a while—Alas! to have to renounce all joy—forever
more!
She fell asleep towards morning, and woke with a start as her alarm
clock thundered. But her face was set like marble, and there was not
a trace of weakness upon it when she appeared at the family
scramble, which did duty for breakfast.
There had been a row between Fred and Gladys, the sister a year
older than herself, who was a saleswoman at a fashionable
dressmaker's establishment. Matilda, the eldest of the family, was
trying to smooth matters while she sewed up a rent in the skirt
which Ethel, the youngest, would presently wear to the school "for
young ladies" which she daily attended. This, the most youthful Miss
Bush, meanwhile sat in a very soiled Japanese quilted dressing
gown, devouring sausages. There were bloaters on the table, too,
and treacle—and the little general servant was just bringing in the
unsavory coffee in the tin coffeepot.
Tea had been good enough for them always in the father's time, and
Matilda for her part could not see why Fred had insisted upon having
coffee, on the strength of a trip to Boulogne on bank holiday.
But there it was! When Fred insisted, things had to be done—even if
one hated coffee!
Katherine Bush loathed most of her family. She had not an expansive
nature, and was quite ruthless. Why should she love them just
because they were her brothers and sisters? She had not asked to
be born among them! They were completely uncongenial to her, and
always had been. It was obviously ridiculous and illogical then to
expect her to feel affection for them, just because of this accident of
birth, so she argued. Matilda, the eldest, who had always been a
mother to the rest, did hold one small corner of her heart.
"Poor old Tild," as she called her, "the greatest old fool living," and
Matilda adored her difficult sister.
How doubly impossible they all appeared now to the unveiled eyes
of Katherine!
"This is simply disgusting stuff, this coffee!" she said, putting her cup
down with a grimace. "It is no more like French coffee than Ett looks
like a Japanese because she has got on that dirty dressing-gown."
"What do you know of French coffee, I'd like to ask—What ho!" Bert,
the brother just younger than herself, demanded, with one of his
bright flashes. "Have you been to 'Boulong for a bit of a song,' like
the Gov'nor?"
"I wish you'd give over calling me the Gov'nor, Bert!" Mr. Frederick
Bush interposed, stopping for a moment his bicker with Gladys.
"Mabel strongly objects to it. She says it is elderly and she dislikes
slang, anyway."
But Albert Bush waved half a sausage on his fork, and subsided into
a chuckle of laughter. He was the recognised wit of the family, and
Ethel giggled in chorus.
Katherine never replied to any of their remarks, unless she wished
to; there was no use in throwing down the gauntlet to her, it
remained lying there. She did not even answer Matilda's tentative
suggestion that she had always drunk the coffee before without
abusing it!
If they only knew how significant the word "before" sounded to her
that morning!
She finished her bit of burnt toast, and began putting on her hat at a
side mirror preparatory to starting. She did not tell Gladys that she
would be late if she did not leave also; that was her sister's own
affair, she never interfered with people.
As she left the dining-room, she said to Matilda:
"I want a fire in my room when I come back this evening, please. I'll
have one every day—Make out how much it will be, and Em'ly's
extra work, and I'll pay for it."
"Whatever do you want that for, Kitten?" the astonished Matilda
demanded. "Why, it is only October yet. No one ever has a fire until
November, even in the drawing-room—let alone a bedroom. It is
ridiculous, dearie!"
"That aspect does not matter at all to me," Katherine retorted. "I
want it, and so I shall have it. I have some work to do, and I am not
going to freeze."
Matilda knew better than to continue arguing. She had not lived with
Katherine for twenty-two years for nothing.
"She takes after father in a way," she sighed to herself as she began
helping the little servant to clear away the breakfast things, when
they had all departed to the West End, where it was their boast to
announce that they were all employed—they looked down upon the
City!
"Yes, it's father, not mother or her family; father would have his way,
and Fred has got this idea, too, but nothing like Kitten's! How I wish
she'd look at Charlie Prodgers and get married and settled!"
Then she sighed again and sat down by the window to enjoy her
one great pleasure of the day, the perusal of the feuilleton in the
Morning Reflector. In these brief moments she forgot all family
worries, all sordid cares—and revelled in the adventures of
aristocratic villains and persecuted innocent governesses and
actresses, and felt she, too, had a link with the great world. She was
a good sound Radical in what represented politics to her, so she
knew all aristocrats must be bad, and ought to be exterminated, but
she loved to read about them, and hear first-hand descriptions of
the female members from Gladys, who saw many in the showrooms
of Madame Ermantine. "Glad knows," she often said to herself with
pride.
Meanwhile, Katherine Bush—having snubbed Mr. Prodgers into
silence in the train—where he manœuvred to meet her every
morning—reached her employers' establishment, and began her
usual typing.
There was work to be done by twelve o'clock in connection with the
renewal of the loan to Lord Algernon Fitz-Rufus—the old Marquis
would be obliged to pay before Christmas time, Mr. Percival
Livingstone said.
Miss Bush, to his intense astonishment, gave a sudden short laugh—
it was quite mirthless and stopped abruptly—but it was undoubtedly
a laugh!
"What is amusing you?" he asked with a full lisp, too taken off his
guard to be as refined and careful in tone as usual.
"The old Marquis having to pay, of course," Katherine responded.
Never once during the whole day did she allow her thoughts to
wander from her work, which she accomplished with her usual
precision. Even during her luncheon hour she deliberately read the
papers. She had trained herself to do one thing at a time, and the
moment for reflection would not come until she could be
undisturbed. She would go back as soon as she was free, to her own
attic, and there think everything out, and decide upon the next step
to be taken in her game of life.
A few burnt sticks, and a lump of coal in the tiny grate, were all she
discovered on her return that evening to her sanctuary. The maid-of-
all-work was not a talented fire-lighter and objected to criticism.
Katherine's level brows met with annoyance, and she proceeded to
correct matters herself, while she muttered:
"Inefficient creature! and they say that we are all equal! Why can't
she do her work, then, as well as I can mine!"
Her firm touch and common sense arrangement of paper and
kindling soon produced a bright blaze, and when she had removed
her outdoor things, she sat down to think determinedly.
She loved Lord Algy—that was the first and most dominant thing to
face. She loved him so much that it would never be safe to see him
again, since she had not the slightest intention of ever drifting into
the position of being a man's mistress. She had tasted of the tree of
knowledge with her eyes open, and the fruit that she had eaten was
too dangerously sweet for continuous food. Love would obtain a
mastery over her if things went on; she knew that she might grow
not to care about anything else in the world but only Algy. Thus,
obviously, all connection with him must be broken off at once, or her
career would be at an end, and her years of study wasted. Even if
he offered to marry her she could never take the position with a high
hand. There would always be this delicious memory of illicit joys
between them, which would unconsciously bias Algy's valuation of
her. She had learned things of consequence which she could not
have acquired in any other way, and now she must have strength to
profit by them. She utterly despised weaklings and had no pity for
lovesick maidens. For a woman to throw over her future for a man
was to her completely contemptible. She probed the possible
consequences of her course of action unflinchingly; she believed so
in her own luck that she felt sure that no awkward accident could
happen to her. But even if this should occur, there were ways which
could be discovered to help her—and since the moment had not yet
come, she would defer contemplating it, but would map out her
plans regardless of this contingency. So she argued to herself.
She could not endure living under the family roof of Laburnum Villa
any longer, that was incontestable; she must go out and learn
exactly how the ladies of Lord Algy's world conducted themselves.
Not that she wished to dawn once more upon his horizon as a
polished Vere de Vere—but that for her own satisfaction she must
make herself his equal in all respects. There had been so many
trifles about which she had felt she had been ignorant, almost every
moment of the three days had given her new visions, and had
shown her her own shortcomings.
"There are no bars to anything in life but stupidity and vanity," she
told herself, "and they at least shall not stand in my way."
The temptation to have one more farewell interview with him was
great, but there was nothing the least dramatic about her, so that
aspect did not appeal to her as it would have done to an ordinary
woman who is ruled by emotional love for dramatic situations; she
was merely drawn by the desire for her mate once more, and this
she knew and crushed.
It would mean greater pain than pleasure to her afterwards, and
would certainly spoil all chance of a career. She gloried in the fact
that she had had the courage to taste of life's joys for experience,
but she would have burned with shame to feel that she was being
drawn into an equivocal position through her own weakness.
Katherine Bush was as proud as Lucifer. She fully understood—apart
from moral questions which did not trouble her—that what she had
done would have been fatal to a fool like Gladys, or to any girl
except one with her exceptional deliberation and iron will. She truly
believed that such experiments were extremely dangerous, and on
no account to be adopted as a principle of action in general. The
straight and narrow path of orthodox virtue was the only one for
most women to follow; and the only one she would have advocated
for her sisters or friends. The proof being that as a rule when
women erred they invariably suffered because they had not the
pluck or the strength to know when to stop.
Katherine Bush was absolutely determined that she should never be
hampered, in her game, by her own emotions or weakness.
Before Lord Algy would return from Wales, she would have left Liv
and Dev's. She had never given him her home address, and there
would be no trace of her. She would look in the Morning Post for
information, and then endeavour to secure some post as companion
or secretary to some great lady. There she would pick up the rest of
the necessary equipment to make herself into a person in whom no
flaws could be found. And when she had accomplished this, then
fate would have opened up some path worth following.
"Some day I shall be one of the greatest women in England," she
told herself, as she looked unblinking into the glowing coals.
Then, having settled her plans, she allowed herself to go over the
whole of her little holiday, incident by incident.
How utterly adorable Algy had been! She found herself thrilling again
at each remembrance—How refined and how considerate! How easy
were his manners; he was too sure of himself, and his welcome in
life, ever to show the deplorable self-consciousness which marked
the friends who came on Sundays, or the bumptious self-assertion of
her brothers, Fred and Bert.
If only she had been born in his world, and had by right of birth
those prerogatives which she meant to obtain by might of
intelligence, how good it would have been to marry him—for a few
years! But even now in her moment of fierce, passionate first love,
which in her case was so largely made up of the physical, her brain
was too level and speculative not to balance the pros and cons of
such a situation. And while she felt she loved him with all her being,
she knew that he was no match for her intellectually, and that when
the glamour faded he would weary her.
But the wrench of present renunciation was none the less bitter—
Never any more to feel his fond arms clasping her—never again to
hear his caressing words of love!
If a coronet for her brow shone at the end of the climb, her heart at
all events must turn to ice by the way, or so she felt at the moment.
He had talked so tenderly about their future meetings. How they
would go again to Paris when he returned from Wales. How she
must let him give her pretty clothes and a diamond ring, and how
she was his darling pet, and his own girl. She knew that he was
growing really to love her; Katherine Bush never deceived herself or
attempted to throw dust in her own eyes. She had eaten her cake
and could not have it. If she had held out and drawn him on, no
doubt she could have been his wife, but it was only for one second
that this thought agitated her. Yes, she could have been his wife—
but to what end? Only one of humiliation. She was not yet ready to
carry off such a position with a certainty of success; she knew she
was ignorant, and that the knowledge of such ignorance would
destroy her self-confidence and leave her at the mercy of
circumstance. So all was for the best. She had not guessed that it
would be so very painful to part from him—dear, attractive Algy! She
could not sit still any longer. A convulsion of anguish and longing
shook her, and she got up and stamped across the room. Then she
put on her outdoor things again and stalked down into the gathering
night, passionate emotion filling her soul.
But when she came back an hour later, after tramping the wet roads
round the common, the battle was won.
And this night she fell asleep without any tears.
 CHAPTER III
It was about a fortnight later that Katherine got Matilda to meet her
at a Lyons' popular café for tea on a Wednesday afternoon.
Livingstone and Devereux had given her a half holiday, being on
country business bent; and having matured her plans, and having
set fresh schemes in train, she thought she might as well
communicate them to the one sister who mattered to her. Matilda
loved an excuse to "get up to town," and had come in her best hat,
with smiling face. Katherine was always very generous to her,
though she was no more careless about money than she was about
other things.
"It is all very well, Tild," she said, in her deep voice, after they had
spoken upon indifferent subjects for a while. "But I am tired of it. I
am absolutely tired of it, so there! I am tired of Liv and Dev—tired of
the hateful old click of the machine with no change of work—I am
tired of seeing the people of another class through the glass screen
—and I mean to get out of it."
"Whatever are you talking of, Kitten!" the elder Miss Bush exclaimed,
as she stirred her cocoa. "Why, Liv and Dev's as good a berth as
you'd get—thirty bob a week, and a whole holiday on Saturday—to
say nothing of off times like this—you must be mad, dearie!" Then
something further in her sister's remark aroused comment.
"And what do you mean by people of 'another class'? Why, aren't we
as good as anyone—if we had their money?"
Katherine Bush put down her empty cup before she replied:
"No, we're not—and if you weren't as ignorant as you are, dear old
Tild, you'd know it. There are lots and lots of classes above us—they
mayn't be any cleverer—indeed, they are often fools, and many
aren't any richer—but they're ladies and gentlemen."
Matilda felt personally insulted.
"Upon my word, Kitten!—If you are such a poor thing that you don't
consider yourself a young lady—I am not. I always did say that you
would pick up rubbishly ideas bothering after those evening lectures
and French classes—instead of coming with Glad and Bert and me to
the cinema, like a decent Christian—it was a low sort of thing to do,
I think, and looked as if we'd none of us had a proper education—
and all they have done for you is to unsettle your mind, my dear—so
I tell you."
Katherine Bush smiled complacently and looked at her sister straight
in the eyes in her disconcerting way, which insured attention. Matilda
knew that she would now have to listen probably to some home
truths. She could manage Gladys very well in spite of her giggles
and irresponsibility, but she had never been able to have the
slightest influence upon Katherine from the moment of their
mother's death, years before, when she had taken her place as head
of the orphaned household. Katherine had always been odd. She
had a vile temper as a child, and was silent and morose, and at
constant war with that bright boy Bert, loved of the other sisters:
Matilda remembered very well many scenes when Katherine had
puzzled her. She was so often scornful and disapproving, and used
to sit there with a book scowling at them on Sundays when a rowdy
friend or two came in to tea, and never once joined in the chorus of
the comic songs they sang, while she simply loathed the
gramophone records.
"You say awfully silly things sometimes, Tild," Katherine announced
calmly. "There would not be any good in my considering myself a
young lady, because at my present stage anyone who really knew
would know that I am not—but I mean to become one some day.
You can do anything with will."
Matilda bridled.
"I don't know what more of a lady you could be than we all are—
Why, Mabel Cawber always says that we are the most refined family
of the whole lot at Bindon's Green—and Mabel ought to know
surely!"
"Because her father was a solicitor, and she has never done a stroke
of work in her life?" Katherine smiled again—it made Matilda feel
uncomfortable.
"Mabel is a perfect lady," she affirmed indignantly.
"I will be able to tell you about that in a year's time, I expect,"
Katherine said, reflectively. "At present, I am not experienced
enough to say, but I strongly feel that she is not. You see, Tild, you
get your ideas of things from the trash you read—and from the
ridiculous nonsense Fred and Albert talk after they come home from
those meetings at the National Brotherhood Club—fool's stuff about
the equality of all men——"
"Of course we are all equal!" broke in Matilda, still ruffled.
Katherine Bush smiled again. "Well, I wish you could see the
difference between Fred and Bert and those gentlemen I see
through the glass screen! They have all got eyes and noses and legs
and arms in common, but everything else is different, and if you
knew anything about evolution, you'd understand why."
"Should I!" indignantly.
"Yes. It is the something inside the head, something in the ideas,
produced by hundreds of years of different environment and a wider
point of view—and it is immensely in the little customs and manners
of speech and action. If you had ever seen and spoken to a real
gentleman, Tild, you would grasp it."
Matilda was quite unmollified and on the defensive.
"You can't have two more honourable, straightforward young fellows
than our brothers in no family in England, and I expect lots of your
gents borrowing money are as crooked as can be!"
Katherine became contemplative.
"Probably—the thing I mean does not lie in moral qualities—I
suppose it ought to—but it doesn't—We had a real sharp last week,
and to look at and to hear him talk he was a perfect gentleman, with
refined and easy manners; he would never have done anything in
bad taste like Fred and Bert often do."
"Bad taste!" snorted Matilda.
"Yes—we all do. No gentleman ever tells people in words that he is
one—Fred and Bert say it once a week, at least. They lay the
greatest stress on it. No real gentlemen get huffy and touchy; they
are too sure of themselves and do not pretend anything, they are
quite natural and you take them as they are. They don't do one
thing at home at ease, and another when they are dressed up, and
they aren't a bit ashamed of knowing anyone. Fred does not speak
to Ernie Gibbs when he is out with Mabel, although they were at
school together!"
"Ernie Gibbs! Why, Kitten, he is only a foreman in the Bindon Gas
Works! Of course not! Mabel would take on!"
Matilda thought her sister was being too stupid!
"Yes, I am sure she would—that is just it——"
"And quite right, too!"
Katherine shrugged her shoulders. There was not much use in
arguing with Matilda, she felt, Matilda who had never thought out
any problem for herself in her life—Matilda who had not the privilege
of knowing any attractive Lord Algys!—and who therefore could not
have grasped the immeasurable gulf that she, Katherine, had found
lay between his class and hers!
"They say Fred is a capable auctioneer because father and
grandfather were—you hear people saying 'it is in the blood'—Well,
why is it, Tild?—Because heredity counts just as it does in animals,
of course. So why, if a man's father and grandfather, and much
further back still, have been gentlemen commanding their inferiors,
and fulfilling the duties of their station, should not the traits which
mean that show as plainly as the auctioneer traits show in Fred
——?"
Matilda had no answer ready, she felt resentful; but words did not
come, so Katherine went on:
"You can't jump straight to things; they either have to come by
instinct through a long line of forebears, or you have to have
intelligence enough to make yourself acquire the outward signs of
them, through watching and learning from those who you can see
for yourself have what you want."
Matilda called for another cup of cocoa—she disliked these views of
Katherine's.
"You see," that young woman went on, "no one who is a real thing
ever has to tell people so in words. Liv and Dev don't have to say
they are two of the sharpest business men in London—anyone can
realise it who knows them. You, and all of us, don't have to tell
people we belong to the lower middle class, because it is plain to be
seen, but we would have to tell them we were ladies and gentlemen,
because we are not. Lord Al—oh! any lord who comes to our office—
does not have to say he is an aristocrat; you can see it for yourself
in a minute by his ways. It is the shams that always keep shouting.
Mabel Cawber insists upon it that she is a tip-top swell; Fred thinks
he is deceiving everyone by telling them what a gentleman he is,
and by not speaking to Ernie Gibbs, who is an awfully good fellow.
Emily says she is a splendid general, and can't even light a fire, and
won't learn how to. George Berker in our office says he is a first-
class clerk, and muddles his accounts. Everything true speaks for
itself. I always mean to be perfectly true, and win out by learning."
Matilda, though somewhat crushed, was still antagonistic.
"I'm sure I hope you'll succeed then, my dear!" she snapped.
"Yes, I shall." Katherine fired her bomb. "It may take me some time,
but that does not matter, and the first step I have already taken is
that I am leaving Liv and Dev's on Friday—and, I hope, going to be
secretary to Sarah Lady Garribardine, at a hundred and ten Berkeley
Square, and Blissington Court, Blankshire!"
"Well, there! You could have knocked me over with a feather!" as
Matilda told Gladys later in the evening. "And wasn't it like Katherine
never telling us a thing about it until everything was almost settled!"
But at the moment, she merely breathed a strangled:
"Oh, my!"
"If I get it, I go to my new situation next week. I had a tremendous
piece of luck coming across it."
"Well, however did you do it, Kitten?" Matilda demanded.
"I saw an advertisement in the Morning Post—it was quite a strange
one, and seemed to be advertising for a kind of Admirable Crichton
—someone who could take down shorthand at lightning speed, and
typewrite and speak French—and read aloud, and who had a good
knowledge of English literature, and thoroughly knew the duties of a
secretary."
"Oh! My!" said Matilda again, "but you can't do half of those things,
Kitten—we none of us know French, do we!"
Katherine smiled; how little her family understood her in any way!
"I wrote first and said they seemed to want a great deal, but as I
had been with Livingstone and Devereux for three years, and
accustomed to composing every sort of letter that a moneylender's
business required, I thought I could soon become proficient in the
other things."
"Well, I never! What cheek!"
"Then I got an answer saying Lady Garribardine liked my
communication, and if I proved satisfactory in appearance, and had
some credentials, she would engage me immediately, because her
secretary, who had been with her for years, had gone to be married
—the salary would be ninety pounds a year with a rise, so it's a
slight move up, anyway, as I am to be kept, and live in the house."
"You are cocksure of getting it, Katherine?"
"Yes—I mean to—I am going to see her on Saturday."
"And what are your references besides Liv and Dev? Some folks
don't like moneylenders."
"I wrote and said I had no others—but they would testify to my
capacity. Liv nearly had a fit when I gave my notice—he almost cried
to get me to stay on. I like the old boy—he is a good sort, and will
tell the truth about me."
"And did they answer?"
"Yes—just to say I was to come for the interview on Saturday."
"They want to see you, anyway—what is the family, I wonder?"
Here Katherine recited the details from Debrett, in which volume she
was very proficient.
"An old lady, then," Matilda commented, "and with no children
except a married daughter! That will be easier for you—but why is
she called 'Sarah'? I often have wondered about that, when I read
names in the Flare. Why 'Sarah Lady Something'—and not plain Lady
Something?"
"It's when the man in possession is married and you are not his
mother," Katherine told her, "and if you are, and still have your
Christian name tacked on, it is to make you sound younger. Dev says
dowagers are quite out of fashion. Every widow is 'Sarah' or
'Cordelia' now in the high society, and when he first went to
business, there were only two or three. Queen Victoria never stood
any nonsense."
Matilda was very interested.
"Whatever will you do about your clothes, Kitten? You have nothing
nobby and smart like Gladys. She could lend you her purple taffeta if
you weren't so tall."
"Oh, I manage all right. I'll have a talk with Gladys to-night; she
sees the right sort of people at Ermantine's, and can tell me what to
get—and I'll buy it to-morrow in my lunch hour."
"Well, I am just rattled," Matilda admitted. "Then you'll be leaving
home quite, dearie?"
"Yes, Tild—and I shan't be sorry except to be parted from you—but I
daresay I shall be able to come and see you now and then."
Matilda looked tearful.
"You never were one of us, Katherine."
"No, I know I never was. I often have wondered what accident
pitchforked me in among you, always the discordant note and the
wet blanket. I hark back to someone, I suppose—I've always
determined to get out, when I was ready."
"You never did care for us—never, Kitten."
Katherine Bush remained quite unmoved.
"No, never for the others—but always for you, Tild—and I'll never
forget you, dear. There, don't be a donkey and cry—the people at
the next table are looking at you."
This argument she knew would calm her sister—who was intensely
sensitive to everyone's opinion.
"And supposing they don't take you?" Matilda suggested, in a still
quavering voice, "and you've given notice to Liv and Dev—I call it
awfully risky."
"Then I will look out for something else—I am determined to make a
change, and see a new world, whatever happens."
After supper that evening, Gladys was invited up to the warmed attic
with Matilda, an honour she duly appreciated. They all stood in
irritated awe of Katherine.
"I want to talk about clothes, Glad," she said, when they neared the
tiny fireplace. "I have told Tild I am going about a new berth on
Saturday."
This caused the same astonishment and exclamations as Matilda had
already indulged in—and when calm was restored, Gladys was only
too pleased to show her superior knowledge.
"I don't want to hear about any of those actresses you dress, or
those ladies who look like them, I want to know what a real, quiet,
well-bred countess, say, would have, Glad."
Miss Gladys Bush smiled contemptuously.
"Oh, a regular frump, you mean—like the ones we can't persuade to
have tight skirts when they are first the fashion, or loose ones when
it changes—that is easy enough—it is to get 'the look' that is
difficult."
"They probably would not engage me if I had 'the look,'" Katherine
remarked cynically.
"You'd better have something like we made for Lady Beatrice
Strobridge last week, then," Gladys suggested. "One of our hands
can copy it at home, but there won't be time by Saturday. You'd
better wear your best blue serge and get a new hat for the first
meeting."
"Lady Beatrice Strobridge must be the Hon. Gerard Strobridge's wife,
my new employer's late husband's nephew. Strobridge is the
Garribardine name." Katherine had looked up diligently the whole
family, and knew the details of each unit by heart.
"She only got married two years ago," Gladys continued. "She was
Thorvil, before—Lady Beatrice Thorvil."
"Wife of the present man's younger brother," quoted Katherine,
remembering Debrett. "He is about thirty-five; the present man is
forty."
"She is a regular dowdy, anyway," Gladys remarked. "One of those—
we have a bunch of them—that wants the things, and yet with their
own touch on them, spoiling the style. They come together
generally, and do make a lot of fuss over each other—calling
'darlings' and 'precious' all the time—fit to make me and the girls die
laughing with their nonsense."
"What is she like—good-looking?" Katherine asked. She only
questioned when she wanted specific information, never idly, and it
was as well to know everything about her possible new employer's
family.
"She would not be bad if she did not stoop so. She hasn't got 'the
walk' neither, no more than the 'look'; sometimes she's all right—at
least, the things are all right when they go home, but she adds bits
herself afterwards, and spoils them."
Here Matilda interrupted.
"Anyway, she is one of the ladies you'll see in your new place, Kitten.
I'd certainly have that same dress, it will just show them you are as
good as they, if you have an Ermantine model."
But Katherine thought differently. She agreed she would have
something in the same subdued style as Lady Beatrice would have
chosen, but not the actual copy, and after settling details the other
two sisters left her for bed.
When they had gone, she sat by the fire and looked deeply into it,
while she thought for a few moments. Then she drew a letter from
her blouse and reread it. It was from Lord Algy. A sweet little love
epistle. Just to tell her he could not possibly wait for the whole
month before seeing her—and was coming up to town the following
week—and would not she lunch with him at the old place—and
perhaps stay with him again at the Great Terminus? It ended with
protestations of passionate devotion.
No—never again—she had tasted of the cup of bliss, and Fate was
asking her to pay no price. She must have courage now to renounce
all further pleasure. Once was an experience, twice would be
weakness—which could grow into a habit—and thence lead to an
abyss which she shuddered to think of.
Katherine Bush had never read Théophile Gautier's masterpiece—but
there was something in her character, as Lord Algy had remarked,
which resembled Mademoiselle de Maupin's.
She went to her little writing-case and got out a sheet of paper, and
then, in her firm round hand which looked like a man's, she wrote
him these few lines:

Dear Algy,
I want you to forget all about me—I loved our little trip, but I
am never going on another. I shall have left Liv and Dev's
before you get back, and you won't see me again. With best
love always.
K. B.

She folded it, put it in the envelope—addressed it and stamped it—

then she put it ready to post in the morning.
Her face was white and set. It takes a strong will to renounce
tangible present happiness, however profound the beliefs in the
future may be.
 CHAPTER IV
Sarah Lady Garribardine said to her nephew, Gerard Strobridge, who
had been lunching with her on that Saturday:
"You must go now, G. I am expecting a new secretary."
"How will you get on without Miss Arnott, Seraphim? I thought she
was irreplaceable."
"So she is—I am interviewing quite a new type—she has been a
moneylender's shorthand typist."
Mr. Strobridge raised his eyebrows—and smiled his whimsical smile.
His Aunt Sarah always was original.
"Then I'll leave you—Beatrice has at last made up her mind not to
chuck the Arberrys, so we motor down at three o'clock."
"Has Beatrice been unusually tiresome?"
"N-no—she has been writing odes all the morning."
"You ought never to have married, G.—You would not have if Alice
Southerwood had not become a widow—a man can't always face his
obvious obligations."
Gerard Strobridge laughed.
"Then I shall kiss your hand and say farewell until next week—wisest
of aunts!"
He suited the action to the word, and left the room just as the butler
was about to open the door and announce:
"Miss Bush, Your Ladyship."
He glanced quickly at Katherine—this was the young person who
would take the estimable Miss Arnott's place, he supposed. She was
quite ordinary looking.—He went on down the stairs.
"Come and sit here in the light, please," Lady Garribardine said, as
Katherine Bush came towards her.
It was a very well-arranged Katherine, in the best blue serge—and a
new hat—not of Gladys' choosing. The mop of hair was twisted tight
without the least pretension to express "the look,"—some grey
suede gloves—bought in Paris by Lord Algy—were on the wonderful
hands which remained perfectly still in their owner's lap.
"How old are you?" asked Lady Garribardine by way of a beginning.
"I was twenty-two last September." There was not a trace of
nervousness in Katherine Bush's deep voice—indeed she felt none.
"And what does your family consist of—what is your status in life?"
Lady Garribardine felt perhaps she ought to ascertain this before
going further.
"We are just middle class. My father was an auctioneer at Bindon's
Green where we live. He and my mother are both dead. I have a
sister who is a saleswoman at Madame Ermantine's, the others are
at home. My eldest brother has taken father's place, the younger
one is in a bank."
"And how long have you been at this business?"
"Since I was nineteen—before that I kept the accounts at a pork
butcher's."
"Indeed!—--And what makes you think you would be capable of
filling my situation?"
"It is not very easy to be a competent moneylender's secretary and
a shorthand writer."
"No—perhaps not."
"Mr. Livingston and Mr. Devereux will tell you that I did not make a
failure of it."
"Really?"
Katherine was silent.
"Really," Lady Garribardine repeated again. "You mean that you
think you can pick up things quickly."
"Yes."
"It is certainly an advantage. I hoped to find something exceptional
when I advertised."
"Yes, I noticed that—and it was because your advertisement was
unusual that I applied for the post."
She rather wondered if she ought to have put in any "Ladyships";
she remembered Hanson, Lord Algy's valet, was very prodigal of
such marks of respect—that is what had deterred her. Liv and Dev
often used them, too—to new and prosperously connected clients—
but she did not wish to be subservient more than was necessary.
She would watch and listen—as she had watched about the oysters.
"Can you read aloud?"
Lady Garribardine was fixing her with her flashing brown eyes, which
contrasted so unfavourably with the bronze-gold wig she wore so
bravely.
"I have never tried. If I did it wrong the first time and you corrected
me, I expect I wouldn't do it twice."
"That is something—and your voice is refined—you did not acquire
that at the—er—pork butcher's?"
"No, I acquired it by listening to members of the upper classes who
came to borrow money—I had a cockney twang like my sisters, I
daresay, in the beginning."
"That shows you can learn things."
"Yes, it is only stupid people who can't."
"You are not stupid, then?"
"No, but Mr. Livingston or Mr. Devereux can tell you; either will speak
for me."
Lady Garribardine was amused; she digressed a little from her cross-
examination.
"You found Jews agreeable to work with?"
"Very. You know where you are with them. They do not pretend, and
they are very generous."
"In-deed!"
"Yes—people have a preconceived notion of Jews, I find—quite
faulty as a rule—they know what to pay for—they are far less fools
than other races. I respect them."
"That is most interesting."
Katherine was silent again.
"Why did you leave them?"—after a pause in which Lady
Garribardine was pitilessly scrutinising her possible secretary.
"Because I had learned all that I could there, and I wanted a new
vista——"
"And you think you would find it with me?"
"With any lady in your world—you can learn things wherever you go,
if you wish to."
"Very true. And how about French—you speak that?"
Katherine Bush reddened a little. A memory came to her of the
profound shock that the French of Paris had been to her ear.
"I can write it quite correctly—but I have discovered that my
pronunciation is ridiculous." She confessed it quite frankly.
"How did that happen?"
"I taught it to myself—mostly—and then I heard it spoken—and I
knew mine would sound wrong."
"Do you think you could overcome that?"
"Yes, if I were in France long enough."
"Have you travelled?"
"No—not really. I have been to Paris for a holiday once—I have only
learnt about places."
"And English literature?"
"It is the thing I care most for—I have read a great many books. I
read usually until about one in the morning."
"Have you a good temper? You are not uppish, eh?"
"I suppose it depends—I know that when you take money to do a
thing you have got to do it, and put up with orders and manners
that you would not stand for one second if you were the person
paying."
"That is quite a good definition of respectful service."
"It is common sense."
"You appear to have some of that."
Again silence.
"I have not a good temper!" Lady Garribardine laughed—she was
greatly diverted.
"I guessed not."
"How?"
"I had to read characters quickly at Livingston and Devereux's——"
"You are observant?"
"I think so——"
"Can you play the piano?"
"I could once, and I had a queer gift for reading the notes—but I
have never practised since we had a gramophone—I grew to loathe
music."
"That is hopeful——"
Then Her Ladyship got up and went to her writing-table, terribly
littered with all sorts of papers. She dived among a conglomerate
mass—and picked up two letters.
"Would you oblige me by answering these, Miss—er—Bush? I could
then better judge of your capabilities."
Katherine took them; on one envelope was written in a spidery hand
in pencil, "Refuse gracefully;" upon the other, "Get out as best can."
She looked for a portion of the blotting pad which was clear enough
to use, then she sat down and selected a pen, while she glanced up
with her steady wise eyes.
"Has Your Ladyship any particular paper for this sort of thing?" Here
was a suitable moment for the use of the honorific she felt.
"Yes, that white paper with the coronet in plain black and the
address."
Lady Garribardine sat down by the fire and stared into it. She had
not been so interested in a specimen of humanity for years.
Katherine Bush read the letters through carefully and the first one a
second time, then she began to write:

To the Secretary of the League for Discouraging Polygamy

among the Mohammedans of India:
Dear Sir,
I am asked by Sarah Lady Garribardine, to tell you that while
sympathising deeply with the admirable object of your League,
she thinks the field over which it must obviously be spread is
too vast for a small contribution to be of much avail, and
therefore, while thanking you for your interesting papers upon
the subject, she is sorry that she is unable to forward you any
more substantial help.
I am, dear sir,
Yours faithfully,
 Katherine Bush (Secretary).

The other letter ran:

To the Matron of the Nonconformist Detention and Penitential

Hostel for Lost Women:
Madam,
I beg to inform you that Sarah Lady Garribardine is leaving town
shortly and therefore cannot avail herself of the pleasure and
honour of visiting your useful institution. She desires me to
express to you her thanks for your invitation.
I remain, madame,
Yours faithfully,
K. Bush (Sec.).

She looked carefully to see what style of address was necessary and
wrote out the envelope—and when all was ready she rose and took
them to the young-old lady by the fire.
She stood quite still while they were perused, and then smiled
inwardly when Lady Garribardine gave a cynical chuckle.
"I think you will do very well, Miss Bush! Please find some stamps,
and put them in that basket to be posted—and—er—you can ring
the bell—I shall expect you—bag and baggage—on Wednesday
next."
This was abrupt, but Katherine Bush felt it was what it should be.
"You do not require the testimony of Mr. Livingston or Mr.
Devereux?"
"No—I can judge for myself—er—Good morning."
The bell had been answered almost instantly and so, bowing,
Katherine Bush followed the servant down the stairs, and soon found
herself in the street, a strange sense of content in her heart.
She knew the West End very well—and walked briskly along Hill
Street and so on past Dorchester House—into the Park. All the
leaves were off the trees. The November day was beautifully fine
and bright and movement was a pleasure.
So the first part of her new game was won at all events.
She reviewed the whole set of impressions she had taken. Firstly,
that the house was a fine one—it had "the look," if houses could be
said to show this quality. That is, it was beautifully kept and filled
with what she guessed from study at the Wallace Collection must be
rare and costly furniture. There were some things she thought ugly
—but "the look" was often ugly, she knew by experience—from
Gladys' verbose descriptions to Ethel and Matilda.
Apart from "the look" it had an air of distinction. It was the abode of
denizens of Lord Algy's world—that was evident. The man she had
met on the threshold of the morning room door was certainly of his
class—and rather nice-looking.
As for her future employer, she was a new specimen to her.
Katherine meant what the French call a type, but she did not know
this expression.
"She is certainly over sixty," she said to herself. "She is a dark
woman naturally, and her hair ought to be grey. The whole thing is
spoilt by that silly golden wig—curled tight like Royalty's. She would
have quite a nice figure for her age if she were not all pushed up by
those old-fashioned corsets. Why had she such big ears and such
red hands for so great a lady? Her rings were buried in fat. The
circulation was evidently wrong somewhere. As for her voice—it was
one of the voices! The female counterpart of the echoes from over
the glass screen—and the manner was quite as casual.
"Just as insolent as I shall be when I hold the same sort of place.
She was born to it—I shall have acquired it—we both when we are
dead will be said to have well filled our situations."
Thus mused Katherine Bush on a November day in Hyde Park—and
turning out of Albert Gate suddenly she almost walked into the arms
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

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

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com