Get Nature Inspired Optimization Techniques For Image Processing Applications Jude Hemanth Free All Chapters
Get Nature Inspired Optimization Techniques For Image Processing Applications Jude Hemanth Free All Chapters
com
https://textbookfull.com/product/nature-inspired-
optimization-techniques-for-image-processing-
applications-jude-hemanth/
https://textbookfull.com/product/nature-inspired-computing-and-
optimization-theory-and-applications-1st-edition-srikanta-patnaik/
textbookfull.com
https://textbookfull.com/product/human-behaviour-analysis-using-
intelligent-systems-d-jude-hemanth/
textbookfull.com
https://textbookfull.com/product/image-processing-and-gis-for-remote-
sensing-techniques-and-applications-second-edition-liu/
textbookfull.com
https://textbookfull.com/product/immigration-and-housing-rents-
evidence-from-german-reunification-1st-edition-kathleen-kurschner-
auth/
textbookfull.com
A Country Nurse: From Wave Hill to rural Queensland and
almost everywhere in between Thea Hayes
https://textbookfull.com/product/a-country-nurse-from-wave-hill-to-
rural-queensland-and-almost-everywhere-in-between-thea-hayes/
textbookfull.com
https://textbookfull.com/product/babies-in-groups-1st-edition-ben-s-
bradley/
textbookfull.com
https://textbookfull.com/product/501-spanish-verbs-christopher-
kendris/
textbookfull.com
https://textbookfull.com/product/comptia-a-core-1-220-1001-and-
core-2-220-1002-exam-cram-2nd-edition-david-l-prowse/
textbookfull.com
Intelligent Systems Reference Library 150
Nature Inspired
Optimization
Techniques for
Image Processing
Applications
Intelligent Systems Reference Library
Volume 150
Series editors
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: [email protected]
Editors
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
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.
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
ix
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
x Contents
S. N. Kumar (&)
Sathyabama Institute of Science and Technology, Chennai, India
e-mail: [email protected]
A. Lenin Fred H. Ajay Kumar
Mar Ephraem College of Engineering and Technology, Elavuvilai, India
e-mail: [email protected]
H. Ajay Kumar
e-mail: [email protected]
P. Sebastin Varghese
Metro Scans and Research Laboratory, Trivandrum, India
e-mail: [email protected]
Keywords Unsupervised learning Clustering Fuzzy C means
FCM-firefly algorithm FCM-artificial bee colony algorithm FCM-cuckoo
algorithm
1.1 Introduction
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
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.
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.
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.
In this chapter, the Fuzzy c-means Clustering algorithm coupled with optimization
technique was proposed for the segmentation of medical images. In the perspective
of image processing, clustering is defined as the grouping of pixels into a cluster
which is similar between them, while dissimilar pixels belong to other clusters. The
concept of clustering is depicted below in Fig. 1.2. The clustering algorithms can be
classified into two groups; Supervised and Unsupervised. The requirement of prior
knowledge termed as training samples is the key concept of the supervised clas-
sifier. Artificial Neural Network (ANN), Naive Bayes Classifier, and Support
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
6 S. N. Kumar et al.
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
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.
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
bðRÞ ¼ b0 ecR
2
Estimate the light intensity of firefly Ii using the objective function f(Y)
end if
end for j
end for i
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.
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
Random documents with unrelated
content Scribd suggests to you:
life, to the shade of the monastery, and now she was to spend the
last fifteen years of her life on the imperial throne. She was of
sterner stuff than Zoe, and the Senators must have concluded that
she alone could check the audacity of the low-born Paphlagonians.
This does not in itself argue any great strength of character in
Theodora. We must remember that there was always a party of
ambitious eunuchs or statesmen behind each of the names that is
put forward by the historian.
When the news of this decision reached Michael, and the crowd
stormed angrily at the gates of the palace, he sent an officer on a
swift vessel to the Princes’ Islands for Zoe. In the palace she was
quickly stripped of her nun’s robe, and clothed in her former
garments. It is clear that Michael’s uncle, Constantine, who was not
without ability, directed the campaign in the palace. Michael was
advised to take Zoe with him into the imperial lodge overlooking the
Hippodrome and show the citizens, who had gathered in the
enclosure, that all was well. The only reply he got was a shower of
stones, arrows and epithets, and, as the chroniclers remark, the
young lion became at once a timid hare, and proposed to run for
shelter to the monastery at Studion, on the Asiatic side. His uncle
prevented him, however, and marshalled the guards in the fore part
of the palace. The battle which followed ended in a complete victory
for the people. Constantine and Michael fled across the water to
Studion, in the early morning of Wednesday in Holy Week, and the
new Empress Theodora was conducted into the palace over the
corpses of some three thousand of the combatants.
The royal sisters, it will be understood, did not fly into each
other’s arms. Theodora had to thank Zoe for eleven years’
confinement, and Zoe herself was very reluctant to share her power
with her younger sister. However, a formal reconciliation was
arranged by the Senators, and the two Empresses sat side by side to
receive the homage of the leading citizens and decide what was to
be done with the late Emperor and his uncle. If there were any who
wondered in what spirit Theodora would wield her power after a
decade of religious life, they were not left long in doubt. Zoe asked
what the will of her advisers was in regard to the fugitives, and such
cries as “Out with their eyes!” and “Crucify them!” rang furiously
through the chamber. Zoe recoiled and pleaded for leniency, but
Theodora, a much better speaker than her sister, sternly ordered the
prefect to see that their eyes were put out. A great crowd crossed
the sea with the officers, and saw Michael, who had hidden under
the altar, and his more stoical uncle dragged from the chapel. The
same crowd had applauded Michael in his procession hardly a week
before; now they stood by with wild delight to see the brutal
sentence carried out. It was 21st April: Michael the Caulker had
reigned for four months.
For a few weeks the imperial sisters ruled their kingdom in
complete harmony and with exemplary zeal. M. Diehl, too lightly
following the censorious Psellus, rates the intelligence and character
of both at a very low level, but that estimate is hardly supported by
the facts. Few Emperors had dared to attack the administrative
corruption of the Empire as Zoe and Theodora attacked it in the first
freshness of their power, and as we have every reason to believe
that they would have continued to attack it. For centuries the State
had been the easy prey of ambitious eunuchs at Court and corrupt
officials in the provinces. Zoe and Theodora issued decrees to the
effect that all injustice must cease and that the law must be
administered with equity. They themselves sat on the highest
tribunal of the city to hear cases, and the sale of offices was strictly
prohibited. The accounts of the late chief minister were examined,
and Constantine, eyeless and shaven, was brought from his
monastery to explain the enormous deficiency. The power of his
family was broken for ever, and the miserable man disclosed that
5300 pounds of gold (nearly a quarter of a million sterling) was
hidden in a cistern in his house. Legates and petitions were heard
with dignity by the royal sisters, and it must have seemed to many
that the Empire had, by this singular adventure, obtained juster and
finer rulers than it had known for many a century. We cannot
discriminate in the joint public action of the sisters, but it is clear
that the strong will and intelligence of Theodora were the chief
power of the administration. How drastically the Empire needed such
a purification may be gathered from the fact that, when the
patriarch Alexis died in the following year, a secret and dishonest
hoard of gold, amounting to more than £100,000, was discovered in
his palace.
This brilliant example of feminine rule might have been expected
to disarm the old Byzantine prejudice against women, but prejudices
of that nature are too deeply rooted to be displaced by facts. The
cry was raised that an Emperor was needed, and Zoe once more
expressed her willingness to marry. The careful chronicler tells us
that her conduct was not necessarily inspired by a carnal feeling—
she was now sixty-two years old—but that she may have feared that
Theodora and her ministers wished to dislodge her. Her age, no less
than the remarkable conditions of her third and last marriage, will
easily persuade us that the motive was political. There were those
who said that, as Theodora had been the chief agent in expelling
Michael, the throne belonged to her alone, and Zoe sought an ally.
The first noble chosen by her was Constantine Delassenus, who had
almost obtained her hand and the throne fourteen years before. But
Constantine, when he was invited to the Court for inspection, proved
so brusque and independent that he was again dismissed. Her next
choice was Constantine Catepano, a handsome officer of the palace,
with whom, in spite of her age, the gossips of the Court already
connected Zoe somewhat too intimately. Constantine, however, had
a wife living, and this lady is said to have poisoned him as soon as
she heard of the proposal to divorce her.
If we may believe the gossipy chronicles, Zoe met the
disappointment with tranquillity, as she had another lover among the
officials of the palace. Constantine Monomachos, a very handsome
and distinguished and dissolute noble, had been exiled from Court to
Mitylene by Michael IV. on the suspicion of intimacy with Zoe, and
had for some years gilded the hours of his distant exile with the
enjoyment of letters, the pleasures of the table and the affection of
a pretty and devoted cousin. When his second wife had died, he had
obeyed the injunction of the Church to refrain from a third marriage
and had been content with the free companionship of the beautiful
Sclerena, a sister of the distinguished noble Romanus Sclerus—a
member, that is to say, of one of the proudest Byzantine families.
She had followed her lover to Lesbos, used her fortune to mitigate
the harshness of his exile, and was living with him at the time when
Zoe recalled him to Court. “Handsome as Achilles,” uniting a
prodigious strength with a singular delicacy and elegance of
appearance, equally devoted to the robust pleasures of the chase
and the enervating delights of love, Constantine Monomachos at
once returned to his place in the heart of the ageing Empress, and
was invited to wed her. He is said to have stipulated beforehand that
the fair Sclerena should be allowed to come to Constantinople, and
Zoe genially consented. They were married, and Zoe entered upon
the last and strangest part of her strange career.
While the sexless Theodora continued to rule the Empire and put
out the eyes of her enemies, while Constantine revelled in the new
and more exquisite luxuries of his position, Zoe seems quietly to
have enjoyed the secure and restful days which her marriage
obtained for her. She still, with her maids, compounded and distilled
the perfumes which were almost her one luxury, but she now paid a
scrupulous attention to her devotions and burned much incense
before the icons. Sclerena at first dwelt apart, and Constantine set
about building a magnificent palace for her, thinly veiling his liaison
with the pretence of going daily to see the progress of the works. As
the citizens smiled at the connexion, and Zoe seemed to be piously
indifferent to it, he became bolder and asked Zoe to allow him to
bring Sclerena to live in the palace. Again Zoe consented, and the
ménage à trois was maintained in the most pleasant harmony. She
gave Sclerena the title of Empress, embraced her, when they met,
with entire goodwill, and showed her such consideration that she
never visited her husband without first ascertaining if he was
disengaged. Constantine occupied the central part of the palace, and
his wife and mistress had apartments on each side.
Although Zoe now approached her seventieth year, she still
retained the freshness of her complexion and had no wrinkles.
Psellus says that a stranger would have been sure that she was still
a young woman. She shared the pleasures of the gay Court, and
made no protest against the frivolous Constantine emptying the
treasury on his mistress. If we may believe implicitly all the details
given by Psellus, there was little delicacy in the fun which enlivened
the gardens or halls—for Zoe disliked the open air—of the sacred
domain. Music and skilful dancing were too fine for his appreciation.
He liked the broader merriment of mimes, and took especial
pleasure in imitations of stammering. His chief entertainers would go
so far as to represent, pantomimically, the chaste Theodora lying
abed in child-birth, and Theodora herself joined in the loud laughter
of Constantine as the man imitated the shrieks which befitted such
an occasion. The months passed very merrily, and the treasury
emptied.
And as the treasury emptied, and the citizens saw their funds
passing into the marvellous palace which Constantine was building
for Sclerena, clouds began to gather over the life of the epicure. One
day, in the year 1044, as he rode with his guards at the head of a
religious procession, a cry broke from the crowd: “We don’t want
Sclerena as Empress, nor to see our lawful mistresses, Zoe and
Theodora, perish on her account.” The cry was a spark to the
spreading discontent, and the small troop of guards were
surrounded by a threatening mob. Fortunately for the Emperor, the
Empresses were watching the procession from the balcony, and they
sent troops to rescue him. Later, a discontented noble led some
Macedonian troops against the city, and encamped opposite the
Blachernæ gate. Constantine disdainfully ordered a chair to be
placed for him outside the gate, in order that he might see, and be
seen by, the rebels. For a time they were content to sing comic
songs about him—of which there must have been a good supply in
the city—then they made a dash and scattered his guards, and could
have penetrated into the city, possibly taken it, if they had not
foolishly retired. On such slender threads did crowns hang in that
singular Empire.
Sclerena relieved the growing discontent by a premature death,
apparently about the year 1045, and the superb palace which had
been intended for Constantine’s mistress was turned into a
monastery. Five years later Zoe closed her long and romantic career,
at the age of seventy. Constantine mourned for her as if she had
been a beloved child, and even pressed the Church to put her on the
list of the canonized; he may have read how St Theodora had won
the aureole largely by her freedom from jealousy. When it was
found, after a time, that some curious fungi had grown about her
monument, he insisted that they were heaven-sent assurances that
Zoe had been admitted at once into the company of the saints. The
Greek Church, however, was not persuaded to add Zoe to its quaint
list of the blessed, and few will reflect on the many events which
reveal her personality to us without admitting that, whether or no
she was guilty of the positive crimes attributed to her, she had little
or no moral feeling.
Constantine found consolation in the charms of a young Alan
princess who was detained as a hostage at Constantinople. The milk-
white skin and fine eyes of the unknown so fascinated him that he
gave her the imperial title and emptied the remainder of the treasury
upon her and the relatives who flocked to share her fortune. He was
by this time a miserable wreck of his former magnificent person, and
could not sit unaided on a horse, but the Court still rang with
laughter and buffoonery. His favourite, a man who had been raised
from the position of street buffoon to that of Court jester, became so
infatuated with his wealth and privileges that he dreamed of
possessing the pretty Alan princess and the purple. He was caught in
Constantine’s bedroom with a drawn sword. The Emperor asked why
he had attempted assassination, and, when the man said that he
had an irresistible passion to see himself in the crown and imperial
robes, burst into laughter and ordered the attendants to put them
on him. He returned to his position, and, to the amusement of
Constantine, made more open love than before to the fair Circassian
mistress. But the Emperor died in 1054, and his mistress returned to
her previous obscurity.
When it was seen that Constantine was failing, a number of the
nobles and officials conspired to put on the throne Nicephorus
Bryennius, but Theodora’s supporters forestalled the plot. They sent
a swift vessel for her and lodged her in the sacred palace before
their opponents could bring Bryennius from Bulgaria, which he
governed. She seems to have been forced out of affairs during the
later years of Constantine, and the sending of a boat implies,
apparently, that she had retired to the suburbs. She was still, in her
seventh decade of life, erect of form and clear in mind, and drastic
punishment was inflicted on the conspirators. She then began again
to control the affairs of the Empire as she had done in conjunction
with Zoe. She personally received ambassadors and heard trials, and
resumed her war on corrupt officials. Psellus is disdainful of her rule,
and unjust to her. The only grave defect we can recognize is that she
put the higher offices and commands at the disposal of men who
were less distinguished for ability than for devotion to her. A very
strong provincial aristocracy had by this time arisen in the Empire,
and from their vast estates a number of able nobles and officers
kept a discontented eye on the hierarchy of eunuchs at
Constantinople.
Theodora, conscious of her vigour, and sustained by the
prophetical assurance of a monk that she would wear the crown for
a long time, maintained her power for three further years, and then
became seriously ill. It is said that she chose an aged and feeble
noble of the city, Michael Stratioticus, to don the purple, but one is
rather disposed to see in the choice of Stratioticus the action of the
Court party, whose influence was threatened by the provincial
nobles. Theodora still confided in the monk’s prophecy; she had the
aged soldier brought to her sickbed and bound him by the direst
oaths to promise obedience to herself. She died a few days later,
however, on 30th August 1057, leaving the crown to the frail charge
of Michael VI. The historian must regret that Theodora had not a
larger opportunity to prove her value as a ruler and exhibit her
personality. She was a woman of great vigour and generally high
political ideals, and she incurs the reproach only of stooping at times
to the common Byzantine level in securing her power. It was not
she, but the contemptible Constantine, who emptied the treasury for
frivolous purposes, and, in spite of the light disdain of Psellus, her
rule compares most favourably with that of most of the Emperors.
CHAPTER XI
EUDOCIA
T
HE struggle which Theodora had foreseen was not long
deferred after her death, and Michael Stratioticus was
compelled, after a few months of feeble imperial
experiment, to retire to the private life from which he had been
unwisely drawn. The great territorial nobles—one might almost say,
the feudal nobles—concentrated upon the capital and put one of
their number, Isaac Comnenus, upon the throne. Isaac had in earlier
years married a Bulgarian princess, and her career as mistress of a
large provincial domain, and then as Empress of Constantinople,
suggests a very interesting study. Unfortunately, her husband’s reign
lasted only two years, and the events yield us only few and fleeting
glimpses of the new Empress.
T
HE distinguished family of the Comneni has already made its
appearance in our narrative. It may be recalled that the last
chapter opened with a march of the great provincial nobles
upon the capital, and the placing of one of their ablest
representatives, Isaac Comnenus, upon the throne. Isaac’s brave life
had ended in heroic foolishness. Terrified by an apparition, he
embraced the monastic life, ignored the natural desire of his brother
John to succeed him, and handed the crown to the Ducas family.
During the reign of Eudocia the widow of John Comnenus, Anna,
remained in Constantinople to guard the fortunes of her children and
eventually to help them to secure the throne. She was a woman of
the old Roman build, rather than Byzantine; strong, ambitious, able
and despotic. The Cæsar John Ducas looked on her with just
suspicion, and accused her of treasonable correspondence with
Romanus, when he was struggling to regain his throne. She boldly
asserted that the letters were forged, and brandished an image of
Christ in the eyes of her judges; but it was expedient to condemn
her, and she passed to the melancholy Princes’ Islands.