Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more ➡
Download
Standard view
Full view
of .
Add note
Save to My Library
Sync to mobile
Look up keyword
Like this
10Activity
×
0 of .
Results for:
No results containing your search query
P. 1
Effective MSE optimization in fractal image compression

Effective MSE optimization in fractal image compression

Ratings: (0)|Views: 1,255|Likes:
Published by ijcsis
The Fractal image compression encodes image at low bitrate with acceptable image quality, but time taken for encoding is large. In this paper we proposed a fast fractal encoding using particle swarm optimization (PSO). Here optimization technique is used to optimize MSE between range block and domain block. PSO technique speedup the fractal encoder and preserve the image quality.
The Fractal image compression encodes image at low bitrate with acceptable image quality, but time taken for encoding is large. In this paper we proposed a fast fractal encoding using particle swarm optimization (PSO). Here optimization technique is used to optimize MSE between range block and domain block. PSO technique speedup the fractal encoder and preserve the image quality.

More info:

Published by: ijcsis on Jun 12, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See More
See less

02/08/2011

pdf

text

original

 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
Effective MSE optimization in fractal imagecompression
A.Muruganandham
Sona College of Technology,salem-05.,India.
 
Dr.R.S.D.wahida banu
Govt Engineering CollegeSalem-11, India
 
 Abstract- The Fractal image compression encodes image at low bitrate with acceptable image quality, but time taken for encoding islarge. In this paper we proposed a fast fractal encoding using particle swarm optimization (PSO). Here optimization technique isused to optimize MSE between range block and domain block. PSO technique speedup the fractal encoder and preserve the imagequality. Keywords- mean square error (MSE) ,particle swarm optimization (PSO), fractal image compression (FIC), Iteration Function System (IFS)
I.
 
I
NTRODUCTION
The idea of the fractal image compression (FIC) is based onthe assumption that the image redundancies can be efficientlyexploited by means of block self-affine transformations. Thefractal transform for image compression was introduced in1985 by Barnsley and Demko [1,2]. The first practical fractalimage compression scheme was introduced in 1992 by Jacquin[3,4]. One of the main disadvantages using exhaustive searchstrategy is the very high encoding time. Therefore, decreasingthe encoding time is an interesting research topic for FIC [3, 4].An approach for decreasing encoding time is using thestochastic optimization methods such as genetic algorithm(GA). Some recent GA-based methods are proposed to improvethe efficiency [5, 6]. The idea of special correlation of an imageis used in these methods, which is of great interesting. Whilethe chromosomes in GA consist of all range blocks which leadsto high encoding speed and particular properties of naturalimages have never been used that will results in lose of visualeffect in the retrieved image.Other researchers focused on improvements of the searchprocess to make it faster by tree structure search methods[12,13], parallel search methods [14,15] or quad treepartitioning of range blocks [9,16,].In this paper, we present a fast fractal encoding usingparticle swarm optimization. The outline of the remaining partof this paper is as follows: Section II includes fractal imagecoding. Section III involves implementation of PSO. SectionIV concerns the proposed fast fractal encoder using PSO, andin Section V experimental results are included. In Section VI,we present our conclusions.II.
 
F
RACTAL
I
MAGE
C
OMPRESSION
A
LGORITHM
 The Iteration Function System (IFS) is the fundamentalidea of fractal image compression in which the governingtheorems are the Collage Theorem and the ContractiveMapping Fixed-Point Theorem [7]. The encoding unit of FICfor given gray level image of size N x N is (N/L)
2
of non-overlapping range blocks of size L x L which forms the rangepool
 R
. For each range block 
v
in
 R,
one search in the (N - 2L +1)
2
overlapping domain blocks of size 2L x 2L which forms thedomain pool
 D
to find the best match. The parametersdescribing this fractal affine transformation of domain block into range block form the fractal compression code of 
v
.The parameters of fractal affine transformation
Φ
of domain block into range block is domain block coordinates-(
tx,ty
)
 ,
Dihedral transformation-
d,
contrast scaling-
 p,
brightnessoffset-
q.
Flowchart for this fractal affine transformation isillustrated in fig. 1 and it is given by
,),(000222101211),(
 
q y x y xu y x paaaa y xu y x
 (1)where the 2 x 2 sub-matrix
22211211
aaaa
is one of theDihedral transformations in (2)
,10010
 
,10011
 
,10012
 
,10013
,01104
 
,01105
 
,01106
 
0
 
1107
(2)
238http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
Fig. 1 fractal affine transformation of domain block intorange block The above parameters are found using the followingprocedure1.
 
the domain block is first down-sampled to L x L anddenoted by
u
 2.
 
The down-sampled block is transformed subject to theeight transformations Tk:k = 0,. . . ,7 in the Dihedralon the pixel positions and are denoted by
u
,
= 0,1, .. . ,7, where
u
0
=
u
. The transformations
1
and
2
 correspond to the flips of 
u
along the horizontal andvertical lines, respectively.
3
is the flip along both thehorizontal and vertical lines.
4
,
5
,
6
, and
7
are thetransformations of 
0
,
1
,
2
, and
3
performed by anadditional flip along the main diagonal line,respectively.3.
 
For each domain block, there are eight separate MSEcomputations required to find the index d such that
:)),((MSEmin{arg
vqu p
}7,...,1,0
 (3)where
 
10,22
)),()0,(( 1),(MSE
 L ji
 jiv jiu  Lvu
 (4)Here,
 p
and
q
can be computed directly as
,),(, ),(),(,
210102101010102
 
 Li L j Li L j Li L j
 jiuuu L  jiv jiuvu L  p
 (5)
.),(),( 1
101010102
 
 Li L j Li L j
jiu p jiv  Lq
 (6)4.
 
As
u
runs over all of the domain blocks in
 D
to findthe best match, the terms
x
and
y
can be obtainedtogether with
and the specific
 p
and
q
correspondingthis
, the affine transformation (1) is found for thegiven range block 
v
.In practice,
x
,
y
,
,
 p
, and
q
can be encoded using log
2
(N),log
2
(N), 3, 5, and 7 bits, respectively, which are regarded as thecompression code of 
v
. Finally, the encoding process iscompleted as
v
runs over all of the (N/L)
2
range blocks in
 R
.Fig. 2 show the MSE vs. quantization parameter for anrandomly selected range block of size 8 x 8 from 256 x 256Lena image. From fig. 2, choosing 5 bits and 7 bits asquantization parameter for scale and offset value is justified.
3456789101146810121416no. of bits for Offset quantization
      M      S      E
 scale bits = 34567891011
 Fig. 2 MSE Vs. quantization parameterTo decode, chooses any image as the initial one and makesup the (N/L)
2
affine transformations from the compressioncodes to obtain a new image, and proceeds recursively.According to Partitioned Iteration Function Theorem (PIFS),the sequence of images will
converge. According to user’s
application the stopping criterion of the recursion is designed.The final image is the retrieved image of fractal coding.III.
 
P
ARTICLE
S
WARM
O
PTIMIZATION
(PSO)PSO is a population-based algorithm for searching globaloptimum. It ties to artificial life, like fish schooling or birdflocking, and has some common features of evolutionary
Partition input image into (N/L)
2
non overlapping rangeblocks of size L x L and (N-2L+1)
2
overlapping domainblocks of size 2L x 2LInput image of sizeN x NFor Range blocks 1 to (N/L)
2
 For Domain blocks 1 to (N-2L+1)
2
 For Orientation 1 to 8Calculate Scaling, Offset andMean Square ErrorStore quantized scaling and offset,orientation and position of the domainblock of minimum MSE
239http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
computation such as fitness evaluation. The original idea of PSO is to simulate a simplified social behavior [8, 9]. Similarto the crossover operation of the GA, in PSO the particles areadjusted toward the best individual experience (PBEST) andthe best social experience (GBEST). However, PSO is unlike a
GA in that each potential solution, particle is ‘‘flying” through
hyperspace with a velocity. Moreover, the particles and theswarm have memory; in the population of the GA memorydoes not exist.Let x
 j,d 
(
) and v
 j,d 
(
) denote the
th
dimensional value of thevector of position and velocity of 
 j
th
particle in the swarm,respectively, at time
. The PSO model can be expressed as
))1(.(.)1()(
,*,11,,
 
 x xcvv
 j j j j
 
)),1(.(.
,#22
 
 x xc
 j
 (7)
),()1()(
,,,
v x x
 j j j
(8)where
*,
 j
 x
(PBEST) denotes the best position of 
 j
th
 particle up to time
-1 and
#
 x
(GBEST) denotes the bestposition of the whole swarm up to time
-1,
φ
1
and
φ
2
arerandom numbers, and
c
1
and
c
2
represent the individuality andsociality coefficients, respectively.The population size is first determined, and thevelocity and position of each particle are initialized. Eachparticle moves according to (7) and (8), and the fitness is thencalculated. Meanwhile, the best positions of each swarm andparticles are recorded. Finally, as the stopping criterion issatisfied, the best position of the swarm is the final solution.The block diagram of PSO is displayed in Fig. 3 and the mainsteps are given as follows:1.
 
Set the swarm size. Initialize the velocity and theposition of each particle randomly.2.
 
For each
 j
, evaluate the fitness value of 
 x
 j
and updatethe individual best position
*,
 j
 x
if better fitness isfound.3.
 
Find the new best position of the whole swarm.Update the swarm best position x
#
if the fitness of thenew best position is better than that of the previousswarm.4.
 
If the stopping criterion is satisfied, then stop.5.
 
For each particle, update the position and the velocityaccording (8) and (7). Go to step 2.Fig. 3 Block diagram of PSOIV.
 
P
ROPOSED
M
ETHOD
 In the proposed fast fractal encoding using PSO, we reducethe encoding time by reducing the searching time to find a bestmatch domain block for the given range block from all domainblocks.Flowchart of the fractal encoding of the proposed method isshown in fig. 4.Fig.4 Fractal encoding of proposed methodDomain block of minimum MSE is found by PSOusing the steps given below:
Partition input image into (N/L)
2
non overlapping rangeblocks of size L x L and (N-2L+1)
2
overlapping domainblocks of size 2L x 2LInput image of sizeN x NFor Range blocks 1 to (N/L)
2
 Store quantized scaling and offset,orientation and position of the domain block of minimum MSECalculate scale, offset, orientation and position of thedomain block of minimum MSE from all domainblocks using PSO
 
Fitness evaluationInitializationExperienceupdatingMovingStoppingcriterionreached?(Start)NoYes(Stop)
240http://sites.google.com/site/ijcsis/ISSN 1947-5500

Activity (10)

You've already reviewed this. Edit your review.
1 thousand reads
1 hundred reads
Bala Gee liked this
smile0life liked this
jrejito liked this
jaydeep2cool liked this
murugasalem liked this
murugasalem liked this
mmdixitmm liked this
arpad_szucs liked this

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->