Review on optimization techniques used for image compression

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 04 Issue: 02 | Feb-2015, Available @ http://www.ijret.org 562
REVIEW ON OPTIMIZATION TECHNIQUES USED FOR IMAGE
COMPRESSION
Shet Reshma Prakash1
, Vrinda Shetty2
1
Student, Computer Science & Engineering, SVIT, Karnataka, India
2
Asst. Professor & HOD, Information Science & Engineering, SVIT, Karnataka, India
Abstract
Image compression is most essential requirement for efficient utilization of storage space and transmission bandwidth. Image
compression technique involves reducing the size of the image without degrading the quality of the image. Currently many image
compression algorithms are used to deal with increasing amount of data involved but still finding the alternative solution is the
area of research. This paper reviews some of the Meta heuristic optimization algorithms used for image compression. These
algorithms are based on swarm intelligence. Swarm intelligence is a relatively new area that deals with the study of behavior
among many entities or objects interacting within the natural or artificial systems. In past few years Swarm Intelligence based
algorithms have been applied to a wide variety of problems in combinatorial and continuous optimization, telecommunications,
swarm robotics, networking, image processing etc. This paper provides an insight of many optimization techniques used for image
compression like Ant Colony Optimization (ACO) algorithm , Harmony Search Algorithm (HSA) and Artificial Bee Colony
algorithm, Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Ant Colony Optimization algorithm is inspired by
the behavior among real ant’s while searching for the food source. Harmony Search Algorithm is inspired by the harmony
improvisation process followed while playing music. Particle swarm optimization is an optimization technique inspired by social
behavior of bird flocking or fish schooling. Artificial Bee Colony algorithm is motivated by the behavior exhibited by honey bees
while searching for the food source. Genetic Algorithm is based on processes observed in the natural evolution.
Keywords: Image compression, Ant Colony Optimization (ACO), Harmony Search Algorithm (HAS), Artificial Bee
Colony (ABC) algorithm, Particle Swarm Optimization (PSO), Genetic Algorithm (GA).
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
Data compression is beneficial because it reduces the
consumption of expensive resources like bandwidth and
hard disk space. Image Compression is a technique which is
concerned with reducing the number of bits required to store
and transmit the image without appreciable loss of
information. There are two types of image compression
namely lossy and lossless. In lossy technique there is loss of
information and hence it is not possible to exactly recover
the original image. Also the quality of the image gets
degraded and compression ratio is high. In lossless
technique no information is lost and hence the original
image can be exactly reconstructed but the compression
ratio is low.
Image compression is most essential solution when the
images are transmitted over the internet or downloaded from
the web pages. There are various techniques for image
compression. JPEG is the most commonly used lossy image
compression technique based on Discrete Cosine Transform
(DCT). JPEG 2000 standard uses Discrete Wavelet
Transform (DWT) which decomposes the image into four
frequency sub bands i.e. LL, LH, HL and HH. In the past
few years many Meta-heuristic Optimization algorithms
have been used to compress the images with minimal loss of
information. There are three types of Meta-heuristic
Optimization algorithms i.e. Physics based, Swarm
Intelligence based and Evolutionary based. Ant Colony
Optimization (ACO), Harmony Search Algorithm (HSA)
and Artificial Bee Colony (ABC) algorithm and Particle
Swarm Optimization (PSO) are swarm intelligence based
Meta-heuristic Optimization algorithms and only Genetic
Algorithm is Evolution based Meta-heuristic algorithm. One
important advantage of using the Optimization technique is
that is helps to increase the compression ratio while
maintaining good quality for the image.
Shu-Chuan Chu et. al [1] and R. C. Eberhert et. al [2] have
discussed about various Swarm Intelligence based
Optimization Algorithms. K. Uma et. al [3] have given the
overview of various fractal image compression algorithms.
Harsha D. Zope et. al [4] have done comparative study for
various image compression techniques like Discrete Cosine
Transform(DCT),Discrete Wavelet Transform(DWT) and
Wavelet Packet Decomposition(WPD). Dervis Karaboga et.
al [5] has discussed the Artificial Bee Colony(ABC)
algorithm in detail. Ryan Rey M. Daga et. al [6], Sarang
Bansod et.al [7] and Z. W. Geem et. al [8] have provided
detailed information on Harmony Search Algorithm(HSA)
and its applications. Simranjeet Kaur et.al [9] have proposed
ACO based algorithm for solving the problems like image
edge detection, image compression and image segmentation.
Christian Blum [10] and C. Martinez [11] have discussed
the use of Ant Colony Optimization (ACO) algorithm for
image compression. M. Mohamed Ismail et.al [12][13] have

_______________________________________________________________________________________
proposed Hybrid Lifting and Adaptive Lifting based Image
Compression using Artificial Bee Colony(ABC) algorithms.
Mitra et. al [14] and Chen Shuwang An Tao Hao Litao [15]
have discussed image compression using Genetic
Algorithm(GA).
The paper is organized as follows. Section 2, 3, 4, 5, 6
provides detailed information of Ant Colony Optimization
(ACO), Harmony Search Algorithm (HSA), Artificial Bee
Colony (ABC) algorithm, Particle Swarm Optimization
(PSO) and Genetic Algorithm (GA). Section 7 and beyond
gives the conclusion and references.
2. ANT COLONY OPTIMIZATION
Ant Colony Optimization(ACO) algorithm was introduced
by Marco Dorigo in 1992 to generate the appropriate
solutions for discrete optimization problems, continuous
optimization problems and some real world problems like
Travelling Salesman problem, routing and load balancing. It
mimics the foraging behavior of real ant colonies. There are
many variants of ACO namely Ant System, Elitist Ant
Colony System, Max-Min Ant System, Rank Based Ant
System etc.
Fig -1: The pheromone trial on the longer path gets evaporated after some duration of time and the entire ant colony follows the
shortest path.
ACO simulates the foraging behavior of the real ants in
which they establish shortest path between the food source
and the nest. While searching for the food source the ant
randomly moves in the area surrounding its nest. While
carrying the food source towards the nest they drop a
chemical pheromone over the ground. Ants can smell
pheromone and evaluate its quantity. Thus larger quantity of
pheromone is concentrated on the shortest path and
pheromone deposited on the longest path begins to
evaporate. The path chosen by the ant contains large amount
of pheromone concentration. Thus after acquiring the food
source the ant leaves pheromone trial based on the quality
and quantity of the food source. This pheromone trial helps
other ants to find the shortest path for the food source. The
indirect exchange of information through pheromone
between the ants is known as stigmergy.
ANT SYSTEM (AS)
It was developed by Dorigo et al. in 1991 to solve the
Travelling Salesman Problem (TSP). In the AS algorithm,
the evaporation occurs all over the trail constructed by each
ant. Besides that, pheromone is deposited along the trail
proportionally to the quality of the corresponding solution.
The pseudo code of the Ant System is as follows

_______________________________________________________________________________________
Ant System (AS)
1. for each ant colony
2. for each ant
3. Generate path
4. Evaluate path
5. evaporate pheromone in trails
6. deposit pheromone on trails
7. end for
8. end for
The ACO meta-heuristic consists of a colony of artificial
ants with the characteristics to search good solutions to
various optimization problems. Artificial ants have been
inspired by the real ones although they can be enriched with
some capabilities which are not found in real ants. Some
similarities with real ants are the existence of a colony of
cooperating individuals, pheromone trail and stigmergy, and
the task of finding the shortest path between the nest and the
food source. The main differences are: (i) artificial ants
move from one discrete state to another, and (ii) they have
an internal memory of the search procedure. Also artificial
ants deposit an amount of pheromone which is a function of
the quality of the solution found.
3. HARMONY SEARCH ALGORITHM
Harmony Search Algorithm (HSA) was developed by Zong
Woo Geem et al. in 2001 and can be applied to solve many
optimization problems like water distribution networks,
ground water modeling, vehicle routing and many more.
Harmony Search is music based metaheuristic optimization
algorithm. When playing the music the orchestrator always
tries establishing harmony by improvisation process. This
harmony in the music is analogous to finding the optimality
in an optimization process. The optimal solution to any
problem corresponds to best solution with given constraints
and parameters. The harmony depends on the aesthetics
quality of music which is determined through pitch
(frequency), timbre (sound quality) and amplitude
(loudness) of musical instrument. Perfect harmony is
achieved by adjusting the pitch which changes the
frequency. In music theory the pitch Pn is expressed using
following formula
Pn = 69 + 12log2 (f/440Hz) (1)
Zong Woo Geem et al. has introduced three methods for the
process of optimization i.e. harmony memory, pitch
adjusting and randomization. The harmony memory consists
of random harmony solutions which are evaluated on the
basis of accepting rate raccept. Pitch adjustment is done by
considering the pitch bandwidth brange and pitch adjusting
rate rpa. So we have following equation
xnew =xold + brange*ε (2)
where xold is the existing pitch value or solution from
harmony memory and xnew is new pitch after pitch adjusting
step. Thus new solution is generated by adjusting the pitch
which is better than existing solution. The pitch adjusting
rate rpa controls the degree of adjustment. Higher value of rpa
causes the solution to scatter around some potential optima.
Hence in many applications we use rpa = 0.1 ~ 0.5.
Randomization deals with the diversity of solutions and
helps to find the global optimal solution. The pseudo code of
Harmony Search Algorithm is
begin
Objective function f(x), x=(x1, x2… xd) T
Generate initial harmonics (real number arrays)
Define (rpa), pitch limits and bandwidth
Define harmony memory accepting rate (raccept)
while( t<Max number of iterations )
Generate new harmonics by accepting best harmonics
Adjust pitch to get new harmonics (solutions)
if(rand> raccept), choose an existing harmonic randomly
else if (rand>rpa), adjust the pitch randomly within limits
else generate new harmonics via randomization
end if
Accept the new harmonics (solutions) if better
end while
Find the current best solutions
end
4. ARTIFICIAL BEE COLONY ALGORITHM
Artificial Bee Colony (ABC) algorithm is swarm
intelligence based optimization algorithm developed by
Dervis Karaboga and colleagues in 2005. It mimics the
intelligent foraging behavior of the honey bees. ABC can be
used for optimizing a large set of numerical test functions.
Three main components of the ABC algorithm are food
sources, employed foragers and unemployed foragers. The
food source is selected by the bee based on some properties
like the amount of nectar, closeness to hive, taste of nectar
and effort required for extracting. Employed forager is the
bee which exploits the food source and provides other bees
with rich set of information like distance, direction and
profitability of the food source. Unemployed forager is the
bee which searches for the food source. They are of two
types i.e. scouts and onlookers. The scouts search the food
source randomly while the onlookers search for the food
source based on information provided by the employed
forager. The exchange of the information takes place in the
dancing area and the related dance is called as the waggle
dance. There is higher probability of selecting the most
profitable food sources by the onlooker bees. Let’s consider
two food sources A and B as shown in Figure2.Initially the
unemployed bee starts searching for the food source without
any previous knowledge and there are two cases. If the bee
is scout it randomly searches for the food source. If the bee
is onlooker it watches the waggle dance and selects the food
source which has highest profitability. After acquiring the
food source the bee becomes employed and starts unloading
the nectar into its hive. After unloading the food the bee
exhibits any one of the following behavior.

_______________________________________________________________________________________
Fig -2: Foraging Behavior of Honey Bees.
1. The bee becomes Uncommitted Follower (UF) by not
returning to the same food source again.
2. The bee returns to dancing area and exchanges the
information to recruit some new bees before returning
to the same food source (EF1).
3. The bee does not return to the dancing area or recruit
new bees but continues to forage the same food
source (EF2).
The main steps of the algorithm are as follows:
1. Population initialization
2. Repeat
3. Employed bees are placed on their food sources.
4. Onlooker bees are placed on food source based
on the amount of nectar associated with a food
source.
5. Scout bee searches randomly for the new food
source.
6. The location of the best food source is
memorized by bee.
7. Until the requirements are met
Initial population has same number of employed bees and
onlooker bees and one scout bee. At the initialization stage
the food sources are randomly selected by the bees and
nectar quantity is evaluated. In ABC algorithm each cycle
consists of three steps as follows.
 In first step the employed bees come into hive and
share the information about the food sources in
dancing area. The onlooker bee gets this information
and searches the most profitable food source. The
employed bee once again returns to the same food
source because its location is memorized in the
previous cycle. If the employed bee encounters a
food source having higher nectar amount it
memorizes this new location and forgets the old
location.
 In second step the onlooker bees select the food
source by evaluating the amount of nectar. As the
nectar amount of the food source increases, the
probability of that food source chosen also
increases.
 In third step when the food source is abandoned it is
replaced by the new food source determined by the
scout bee.
5. PARTICLE SWARM OPTIMIZATION
It is evolutionary computation algorithm which was
developed by Kennedy and Eberhert in 1995. It simulates
the behavior of bird flocking together to seek the food
source. While searching for the food source the birds
communicate with each other and reach target within
minimal duration of time. The population of PSO is

_______________________________________________________________________________________
initialized with random solutions called particles. Each
particle has a capability to freely fly through the search
space with some velocity which can be dynamically
adjusted during search process. The original PSO algorithm
is stated through following equation.
Vmd = Vmd + c1Rand1 () (Pmd – Xmd) + c2Rand2 () (Pgd-
Xmd)…….. (3)
Xmd = Xmd + Vmd ………………………..……………. (4)
Where c1, c2 are positive constants and Rand1 () and Rand2
() generate random numbers between range 0 and 1. Xm1,
Xm2 …..Xmd represents the particles, Pm1 , Pm2 ….Pmd
represents the best fitness value of the particles. The index g
belongs to best particle among the population and Vm1 , Vm2
……Vmd represents the velocities of the particle. Equation
(3) represents how the velocity is dynamically adjusted over
the search process and Equation (4) represents how position
is updated for each flying particle. If the value of Vmd
exceeds user defined threshold Vmax then velocity of all
particles is dynamically adjusted to Vmax. The pseudo code
of the algorithm is given below
1. Initialize the particles in population with some
random velocities and positions.
2. Repeat
3. For each particle calculate the best fitness value.
4. Compare the particles best fitness value with pbest.
If current value is better than pbest then set pbest as
best fitness value and change the location
accordingly for this particle.
5. Index of particle with best fitness value among
whole population is assigned to g.
6. Dynamically change the velocity and position
according to equation (1) and (2).
7. Until termination criterion is met.
6. GENETIC ALGORITHM
Genetic Algorithm was developed by Goldberg and inspired
through Darwin's theory of Evolution. In this algorithm the
initial population consists of chromosomes which represent
solutions to the optimization problem. The fitness value is
calculated for each chromosome which is composed of
multiple genes. The chromosomes with higher fitness value
are selected for crossover operation thereby producing
offspring’s. These offspring’s have combination of parents
genes. Also few chromosomes undergo mutation on genes.
The number of chromosomes which will undergo crossover
and mutation is controlled by crossover rate and mutation
rate value.
Chromosome in the population that will remain for the next
generation are selected based on Darwinian evolution rule.
The chromosome which has higher fitness value will have
greater probability of being selected again in the next
generation. After several generations, the chromosome value
will converge to a certain value which is the best solution
for the problem.
The Algorithm
In the genetic algorithm process is as follows
1. Determine the number of chromosomes, generation,
and mutation rate and crossover rate value
2. Initialize the population with Chromosome.
3. Process steps 4-8 until the number of generations is
met
4. Evaluation of fitness value of chromosomes by
calculating objective function
5. Chromosomes selection
6. Crossover operator
7. Mutation operator
8. Produce Offspring
9. Chromosome with higher fitness value is best
solution.
Fig -3: Flowchart of Genetic Algorithm.

_______________________________________________________________________________________
7. CONCLUSION
In this paper we have reviewed many optimization
techniques which have been used for image compression. In
spite of existence of various image compression algorithms
optimization techniques are widely used by many
researchers to get best compressed image with minimal loss
of information. One important advantage of using the
optimization technique is that it reduces the time of
compression and also helps to increase the compression
ratio without degrading the quality of the image.
REFERENCES
[1] Shu-Chuan Chu, Hsiang-Cheh Huang, John F.
Roddick, and Jeng-Shyang Pan - Overview of
Algorithms for Swarm Intelligence, International
Conference on Computer and Computational
Intelligence(ICCCI 2011) , Part I, LNCS 6922, pp.
28 41, 2011. © Springer-Verlag Berlin Heidelberg
2011.
[2] R.C Eberhart, Y. Shi, J. Kennedy- Swarm
Intelligence, Morgan Kaufmann, 2001.
[3] K.Uma, P.Geetha, A.Kannan, K.Umanath - Image
Compression Using Optimization Techniques, e-
ISSN: 2278-067X, p-ISSN: 2278-800X, Volume 5,
Issue 5 (December 2012), PP. 01-07,
www.ijerd.com
[4] Mrs. HarshaD.Zope, Prof. Mrs.Pallavi Y. Patil -
Comparative Study of Different Approaches of
Image Compression,(ISSN 2250-2459, Volume 2,
Issue 1, January 2012), Website: www.ijetae.com .
[5] DervisKaraboga, BahriyeAkay- A comparative
study of Artificial Bee Colony algorithm, Applied
Mathematics and Computation, Volume 214, 2009,
pp 108-132, Website :
www.elsevier.com/locate/amc.
[6] Ryan Rey M. Daga, John Paul T. Yusiong- Image
compression using Harmony Search Algorithm,
International Journal of Computer Sciences, Vol 9,
Issue 5, No 3, 2012, www.IJCSI.org.
[7] Sarang Bansod, Sweta Jain- Improved Harmony
Search Algorithm for Color Image Compression,
International Journal on Recent and Innovation
Trends in Computing and Communication, Volume
2, Issue 6, June 2014.
[8] Z. W. Geem- Music inspired Harmony Search
Algorithm Theory and Applications, Studies on
Computational Intelligence, Vol 199, 2009, pp. 2-
13.
[9] Simranjeet Kaur, Prateek Agarwal, Rajbir Singh
Rana - Ant Colony Optimization : A Technique used
for Image Processing, IJCST Vol. 2, Issue 2, June
2011.
[10] Christian Blum- Ant Colony Optimization-
Introduction and Recent Trends, Physics of Life
Reviews, Vol2,2005, pp 353-373,
www.elsevier.com/locate/amc.
[11] C. Martinez- An ACO Algorithm for Image
Compression, CLEI Electronic Journal, Vol 9, No 2,
2006, pp. 2-5.
[12] M. Mohamed Ismail, Dr.K.Baskaran - Hybrid
Lifting Based Image Compression Scheme Using
Particle Swarm Optimization Algorithm And
Artifical Bee Colony Algorithm,International
Journal of Advanced Research in Computer and
Communication Engineering Vol. 3, Issue 1,
January 2014.
[13] M. Mohamed Ismail, Dr. K. Baskaran - Adaptive
Lifting Based Image Compression Scheme Using
Artificial Bee Colony Algorithm, International
Journal of Electronics Communication and
Computer Engineering, Volume 4, Issue 1, ISSN
(Online): 2249–071X, ISSN (Print): 2278–4209.
[14] S.K Mitra, C. A Murthy, M.K. Kundu- Technique
for fractal image compression using Genetic
Algorithm, IEEE Trans, Image Processsing, Vol 4,
No. 7, pp. 586-593, 2006.
[15] Chen Shuwang An tao Hao litao, IEEE 2009, DCT
based image compression using Genetic Algorithm.
BIOGRAPHIES
Shet Reshma Prakash received B.E in
Information Technology from Mumbai
University and currently pursuing M-
Tech in Computer Science from VTU,
Belgaum. Field of interest includes
image processing.
Vrinda Shiva Shetty received B.E
from Gulbarga University and M.Tech
degree from VTU in Computer Science
and Engineering and presently pursuing
Ph.D. in Image Compression from the
Gulbarga University Gulbarga,
Karnataka. Field of Interest includes
Image Processing, Evolutionary Computation.

Review on optimization techniques used for image compression

More Related Content

Viewers also liked (14)

Similar to Review on optimization techniques used for image compression (20)

More from eSAT Journals (20)

Recently uploaded (20)

Review on optimization techniques used for image compression