research article medical image visual appearance improvement using...

Research ArticleMedical Image Visual Appearance ImprovementUsing Bihistogram Bezier Curve Contrast EnhancementData from the Osteoarthritis Initiative

Hong-Seng Gan1 Tan Tian Swee2 Ahmad Helmy Abdul Karim3

Khairil Amir Sayuti3 Mohammed Rafiq Abdul Kadir4

Weng-Kit Tham5 Liang-Xuan Wong5 Kashif T Chaudhary6

Jalil Ali7 and Preecha P Yupapin8

1 Department of Biotechnology and Medical Engineering Faculty of Biosciences and Medical EngineeringUniversiti Teknologi Malaysia 81310 Skudai Johor Malaysia

2 Department of Biotechnology and Medical Engineering Medical Device and Technology GroupMaterial and Manufacturing Research Alliance Faculty of Biosciences and Medical Engineering Universiti Teknologi Malaysia81310 Skudai Johor Malaysia

3 Department of Radiology School of Medical Sciences Universiti Sains Malaysia 16150 Kubang Kerian Kelantan Malaysia4Medical Device and Technology Group Faculty of Biosciences and Medical Engineering Universiti Teknologi Malaysia81310 Skudai Johor Malaysia

5 Department of Control Engineering and Mechatronic Engineering Faculty of Electrical EngineeringUniversiti Teknologi Malaysia 81310 Skudai Johor Malaysia

6 Institute of Advanced Photonics Science Nanotechnology Research Alliance Universiti Teknologi Malaysia81310 Johor Bahru Malaysia

7 Institute of Advanced Photonics Science Nanotechnology Research Alliance Universiti Teknologi Malaysia81310 Skudai Johor Malaysia

8Department of Physics Advanced Studies Center Faculty of Science King Mongkutrsquos Institute of Technology LadkrabangBangkok 10520 Thailand

Correspondence should be addressed to Tan Tian Swee tantsweebiomedicalutmmy

Received 11 February 2014 Accepted 19 April 2014 Published 20 May 2014

Academic Editor Ngaiming Kwok

Copyright copy 2014 Hong-Seng Gan et alThis is an open access article distributed under theCreative CommonsAttribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Well-defined image can assist user to identify region of interest during segmentation However complex medical image is usuallycharacterized by poor tissue contrast and low background luminance The contrast improvement can lift image visual quality butthe fundamental contrast enhancement methods often overlook the sudden jump problem In this work the proposed bihistogramBezier curve contrast enhancement introduces the concept of ldquoadequate contrast enhancementrdquo to overcome sudden jump problemin knee magnetic resonance image Since every image produces its own intensity distribution the adequate contrast enhancementchecks on the imagersquos maximum intensity distortion and uses intensity discrepancy reduction to generate Bezier transformcurve The proposed method improves tissue contrast and preserves pertinent knee features without compromising natural imageappearance Besides statistical results from Fisherrsquos Least Significant Difference test and theDuncan test have consistently indicatedthat the proposed method outperforms fundamental contrast enhancement methods to exalt image visual quality As the study islimited to relatively small image database future works will include a larger dataset with osteoarthritic images to assess the clinicaleffectiveness of the proposed method to facilitate the image inspection

Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 294104 13 pageshttpdxdoiorg1011552014294104

2 The Scientific World Journal

1 Introduction

Magnetic resonance (MR) imaging allows direct visualizationof knee cartilage and quantitative measurement on cartilageto monitor osteoarthritis (OA) progression [1] The advance-ment of MR imaging contributes greatly to the search foreffective OA imaging biomarker [2] Imaging biomarker isdefined as ldquoany anatomic physiologic biochemical ormolec-ular parameter detectable with one ormore imagingmethodsused to diagnose the presence andor severity of disease [3]rdquoSince imaging biomarker is crucial for developing diseasemodifying osteoarthritis drug (DMOAD) [4] the NationalInstitutions of Health (NIH) has launched the OsteoarthritisInitiative (OAI) in a public-private consortium To date atotal of 4796 participants have been recruited from four clini-cal institutions (University of Maryland School of MedicineBaltimore Md Ohio State University Columbus OH Uni-versity of Pittsburgh Pittsburgh Pa Memorial Hospital ofRhode Island Pawtucket RI) to provide comprehensiveanalysis for OA progression and identification of suitable OAbiomarkers

There have been growing interests in interactive segmen-tation recently [5ndash8]Medical images such as kneeMR imagesare usually too complex for fully automated segmentation toproduce satisfactory results On the other hand interactivesegmentation reflects user intention because the segmenta-tion is initiated according to minimal user intervention inthe forms of scribbles [9] or bounding box [10] Thereforeunnecessary segmentation error such as oversegmentationreported in automated segmentation [11] can be avoidedTypically interactive segmentation can be divided into twogroups boundary based and contour based Although bothgroups present different segmentation approaches initializa-tion of these methods will greatly depend on human knowl-edge Therefore direct incorporation of human knowledgeto commence the segmentation translates into underlyingneed for human interpretation of MR image Distinct featuremanifestation and excellent visual quality of the medicalimage for instance play important role to assist user ininterpreting medical images and then inserting seeds ontoregion of interests

The visual quality of knee MR image can be deterioratedby several phenomena insignificant tissue contrast com-plex knee joint structure and low background luminanceDefinitive tissue contrast facilitates the image interpretationprocess Unfortunately mediocre tissue contrast has beenobserved in knee MR image For example tissue contrastbetween cartilage surrounding muscle tissues and imagebackground only differ slightly As a result poor contrastdifference observed in the image can easily lead to inter- andintraobserver ambiguity during inspection Besides humanknee structure is anatomically complex The knee compart-ment is filled with various types of cartilages knee bonesmuscle fat tissue and ligaments attempting to interpretMR image equipped with poor tissue contrast is laboriousFurther MR image of knee is characterized by large numbersof low intensity background pixels Dark background hard-ens any effective identification of pertinent image featuresEventually above mentioned phenomena give rise to the

prominence of a tissue contrast enhancement model toelevate image visual quality

Contrast enhancement is defined as a remapping processto transform the imagersquos intensity distribution so the imageintensity range can be fully exploited [12] In particularcontrast enhancement has been implemented to extractsalient information protrude image features and improvethe imagersquos visual quality Applications of various contrastenhancement models on medical images for further analysishave been reported in previous studies Ismail and Sim [13]applied dynamic contrast enhancement on low resolutionbrain MR images so important brain image features couldbe protruded Chai et al [14] proposed the use of multipur-pose contrast enhancement technique to overcome the poorcontrast problem reported in the ossification sites of childrenhand bones In the context of knee MR image adequatecontrast enhancement to ameliorate the image appearancewill be meaningful enough for the subsequent interactiveknee cartilage segmentation to succeed

Direct contrast enhancement through traditional his-togramequalization [12] cannot be implemented intomedicalimage processing Traditional histogram equalization distortsthe brightness enhancement by invariably shifting outputmean brightness to middle gray level The resultant imagetherefore is over-enhanced and any brightness sensitiveimage feature is destroyed Besides traditional histogramequalization combines any neighboring gray levels with lightprobabilistic density into one gray level widening the gapwith neighboring gray levels which have heavy probabilisticdensity The flawed intensity redistribution contributes tosevere detail image loss and sudden jump in cumulativedensity function

Fundamental contrast enhancement betterments aredeveloped based on the idea of histogram partitioning Thegist of these improved models concentrates on preservingmean brightness of original image using different intensitythreshold values that is mean and median As such oneof the earliest bihistogram equalization methods brightnesspreserving bihistogram equalization (BBHE) [15] has beendeveloped based on this concept BBHE partitions the globalhistogram into two subhistograms according to mean inten-sity value Then traditional histogram equalization is imple-mented independently onto the subhistograms Given thatBBHE partitions the intensity distribution into two whilemean brightness of input image is preserved Conservation ofmean brightness is deemed imperative in preventing resultantimage from severe intensity distortion

In 1999 Wang et al [16] introduced dualistic subimagehistogram equalization (DSIHE) This method uses medianintensity value to decompose original imagersquos histogramYu postulates that maximum segmentation entropy can beobtained when two subimages have equal areas DSIHEis capable of preserving original imagersquos mean brightnessas well as attaining the most image entropy Meanwhileminimum mean brightness error bihistogram equalization(MMBEBHE) [17] has developed based on iterative selec-tion of optimal threshold value For instance the optimalthreshold valuemust satisfy minimum absolute mean bright-ness error (AMBE) between output histogram and input

The Scientific World Journal 3

histogram Hence MMBEBHE overcomes the problem ofunnatural enhancement and maximizes the degree of bright-ness preservation consequently it downplays the creation ofunwanted artifacts HoweverMMBEBHE is relatively unsuit-able in real image processing because generating AMBEfor each possible input intensity level is computationallycomplicated

Chen and Ramli [18] proposed an iterative extensionof BBHE known as recursive mean separate histogramequalization (RMSHE) RMSHE decomposes original imagehistogram into several subimages based on mean value andperformed traditional histogram equalization up to 119903 timesRMSHE claims to demonstrate better brightness preservationwhen the number of iterations increases Instead highernumbers of iterations reduce the enhancement effect to pro-duce highly identical output imageThe iterative extension ofDSIHE recursive subimage histogram equalization (RSIHE)has been developed by Sim et al [19] Unfortunately RSIHEshares similar problem with RMSHE Furthermore bothrecursive HE models have computational power up to 2119903If higher degree of histogram decomposition is chosen theprocess would become computationally complicated

In this paper a new contrast enhancement model knownas bihistogram Bezier curve contrast enhancement (BBCCE)is proposed The intention of BBCCE is to improve imagevisual quality through appropriate contrast enhancementwithout compromising the conservation of salient kneefeatures As such transformation of intensity distributionis performed using remapping process based on Beziertransform curve instead of conventional cumulative densityfunction Besides the important property of bihistogramequalization is retained by partitioning the histogram intotwo subhistograms to curtail predominance of low intensitybackground pixels

2 Materials and Methods

21 MR Image Acquisition Twenty healthy baseline (Dataset 0C2) dual-echo steady-state (DESS) knee MR imageswith water excitation (we) [20] from the OAI dataset areselected MR images are acquired using 30 T MRI scan-ner (Siemens Magnetom Trio Erlangen Germany) withquadrature transmit-receive knee coil (USA InstrumentsAurora OH) In this experiment the DESSwe MR imageshave section thickness of 07mm and in-plane resolu-tion of 0365 times 0456mm Other technical parametersare outlined as follows field of view = 140 times 140mmflip angle = 25∘ TRTE = 16347m sec matrix size =384 times 384mm bandwidth = 185Hzpixel Further informa-tion related to the OAI dataset could be found throughhttpoaiepi-ucsforgdatareleaseAboutasp

22 Contrast Enhancement Model Overview Conventionaltransform curve is derived from cumulative density functionThe objective is to stretch the original intensity distributionusing transform curve to cover full dynamic range of theimage Thus imagersquos contrast can be modified Howevercumulative density function based remapping process does

not always yield the desired effect because distortion by dom-inant intensity levels (as indicated by arrow in Figure 1(b))could easily induce sudden jump in conventional transformcurve (as indicated by arrows in Figure 1(c)) resulting inabrupt rise of transformed intensity values On the groundof above mentioned concerns it is believed that an adequatetissue contrast improvement ismost appropriate approach forDESSwe knee MR image

The term ldquoadequaterdquo implies dynamic adjustment oftransformed intensity values to refine the traditional trans-form curve In addition we must preserve prominent imagefeatures and maintain natural visual appearance simultane-ously Hence the proposed BBCCE uses the largest intensitydiscrepancy value deduced from intensity difference curve(Figure 1(d)) to identify furthest intensity distortion causedby traditional histogram equalization This model thereforecould handle arbitrary intensity variability exhibited by dif-ferent images Then global extremaextremum is computedfrom the curve and used as control points to generate Beziertransform curves Finally the Bezier curves are concatenatedto form global transform curve (Figure 1(e)) BBCCE isdeveloped by using MATLAB (Mathworks Natick MA)

23 Image Decomposition MR image is defined with 119871

discrete gray levels as 119883 = 119883(119894 119895) where 119883119896(119894 119895) isin

119883119900 1198831 119883

119871minus1 By treating these pixel intensities as ran-

dom variables the probability density function is estimatedas 119901(119883

119896) for MR image Probability density function depicts

how frequent certain pixel intensity occurs Hence proba-bility density is generated by dividing the pixel occurrenceof certain gray level 119896 119899

119896over the total number of pixels 119899

Consider the following

119901 (119883119896) =

119899119896

119899

(1)

In this work two subhistograms are produced after his-togram decomposition The partition is intended to confineany abrupt distribution skew caused by large amounts ofbackground pixels to lower histogram To compute histogramdecomposition the mean MR image pixel intensity 120583 iscalculated Formulation for the arithmetic mean is illustratedas follows

120583 =

sum119896

119896=0119899119896(119883119896)

119899

=

119896

sum

119896=0

119901 (119883119896)119883119896 (2)

24 Intensity Difference Curve Histogram of MR imagecan be derived from the plot of 119899

119896against 119883

119896 To attain

contrast enhancement through histogram equalization basicidea is to stretch the input histogram to resemble uniformdistribution regardless of the initial histogram shape [12]However pixel intensity is presented in discrete domain forreal image processing and the output probability densityfunction could only get as close as possible to the uniformdistribution During the implementation of histogram equal-ization cumulative density function plays nontrivial roleto remap the pixel intensity distribution The conventional


(a)

0 50 100 150 200 250 3000

1000

2000

3000

4000 PDF

Intensity

Num

ber o

f pix

els

0

1000

2000

3000

4000 Lower PDF

Num

ber o

f pix

els

0 50 100 150 200 250 3000

500

1000

1500 Upper PDF

Intensity

Num

ber o

f pix

el

0 50 100 150 200 250 300Intensity

(b)

0 50 100 150 200 250 3000

50

100

150

200

250

300 CDF

Intensity of pixel

Cum

ulat

ive f

requ

ency

of p

ixels

(c)

0

k

l

50 100 150 200 250 300

0

20

40

60

80

100

120

L

(59 0)

(44 5)

(125 107)

(17 minus11)minus20

ΔL

(d)

0 50 100 150 200 250 3000

50

100

150

200

250

300

Original intensity

Enha

nced

inte

nsity

(e) (f)

Figure 1 Flow of BBCCE computation using knee MR image (a) Original MR image (b) Upper diagram shows histogram of original MRimage where black arrow indicates dominant background intensities and lower diagrams show the decomposition of histogram into lowerand upper subhistograms in BBCCE (c) Cumulative density function of original MR image where black arrows indicate large intensitydistortion which contributes to sudden jump issue (d) Intensity discrepency curve duduced from cumulative density function where upwardand downward red arrows indicate global maximum and global minimum in lower histogram while upward black arrow indicates globalmaximum in upper histogram Leftward black arrow defines the boundary for global extremum in intensity discrepency curve (e) Beziertransform curve generated using control points duduced from intensity discrepency curve (f) BBCCE enhanced MR image


cumulative density function as 119888(119909) and traditional transformfunction as 119891(119909) can be defined as

119888 (119909) =

119896

sum

119896=119900

119901 (119883119896)

119891 (119909) = 119883119900+ (119883119871minus1

minus 119883119900) 119888 (119909)

(3)

As shown in (3) traditional transform function reliesheavily on conventional cumulative density function withoutconsidering the sudden jump issue Sudden jump in conven-tional cumulative density function arises when predominatepixel intensity instigates steep increase in the functionConsequently steep increment over small intensity rangeinduces excessive brightness lift This issue requires seriousattention in the case of knee MR image since large amount ofbackground pixels can skew conventional cumulative densityfunction

In proposed method the solution for this issue is tosmooth the conventional cumulative density function basedon intensity distortion caused by sudden jump Noteworthythe degree of contrast enhancement depends on absoluteintensity difference (AID) which is defined as

AID = 1199091015840

minus 119909 (4)

where 1199091015840 indicates the transformed intensity and 119909 is theoriginal intensity The degree of contrast enhancement couldbe decreased by lowering the AID

The intensity difference curves are deduced by computingdiscrete AIDs for both subhistograms Notably intensity dif-ference curve for eachMR image is distinctThis novel featurein the proposed BBCCE allows us to consider arbitraryintensity distribution variation for smoothing conventionalcumulative density function Thus the transform curveis based on two factors input intensity distribution andresultant intensity fluctuation Suppose that the intensitydifference curve is defined as 119910 = 119868(119909) Then intensitydiscrepancy value IDV isin R IDV illustrates the intensitydiscrepancy for intensity level that corresponds to the criticalpoint where its gradient is equal to zero119889119910119889119909 = 0Within the predefined range several critical points couldexist potentiallyThese critical points119875(119909) as localminimumcan be identified if 11988921199101198891199092 gt 0 and local maximumif 11988921199101198891199092 lt 0 The critical points are subjected to self-imposed boundary condition 119910 = 0 where local minimumis only defined in region 119910 lt 0 and local maximum is definedin region 119910 gt 0 Thus global extrema could be found amongthese critical points by searching for the largest IDV underpredefined constraints

119875Global maximum (119909) = Argmax (IDV) 119910 gt 0

0 otherwise

119875Global minimum (119909) = Argmin (minusIDV) 119910 lt 0

0 otherwise

(5)

Given that the proposed method defines MR imagewithin [0 119871 minus 1] first search attempt is performed in the

range of 0 to 120583 (lower histogram) and 120583 to 119871 minus 1 (upperhistogram) for second search attempt Eventually two sets ofglobal extremum or global extrema are obtained

25 Bezier Transform Curve Bezier curve is popular incomputer graphics and computer-aided application Theparametric curve representing the special case of 119861-splinehas been instrumental for widely diversified applicationsspanning from industrial shape design to game developmentInherently Bezier curve is characterized by convex hull Thecurve is contained inside control polygon guaranteeing thatthe generated curve will not derail off its control polygon Inspite of its versatility and advantages to our best knowledgemultiple Bezier curves have never been applied in medicalimage processing Here in after Bezier curves are introducedto produce smooth remap curve that yields smaller AIDDenote the control points as 119875

0 1198751 119875

119899and is bounded

within the domain 119905 isin [0 1] Bezier curve can be define as[21]

119876 (119905) =

119899

sum

119894=0

119861119899119894(119905) 119875119894 (6)

Bezier curve uses Bernstein polynomial 119861119899119894(119905) as its basis

function in which definition of the polynomial is formulatedin (7) Consider the following

119861119899119894(119905) = (

119899

119894) 119905119894

(1 minus 119905)119899minus119894

(7)

where the binomial coefficient ( 119899119894) is given as

(

119899

119894) =

119899

119894 (119899 minus 1)

if 0 le 119894 le 119899

0 otherwise(8)

Based on intensity difference curve Bezier curve ofsecond degree or third degree is likely to be generated Ifa pair of global extrema is detected from the curve Beziercurve of third degree 119899 = 3 can be deduced with four controlpoints 119875

119894isin 1198750 1198751 1198752 1198753 On the other hand Bezier curve of

seconddegree 119899 = 2with three control points119875119894isin 1198750 1198751 1198752

is selected if global extremum is detectedThe point selectionis automated in ourmethod Specific second and third degreeBezier curves are expressed as follows

119876 (119905) =

2

sum

119894=0

(

2

119894

) 119905119894

(1 minus 119905)2minus119894

119875119894 119899 = 2

3

sum

119894=0

(

3

119894

) 119905119894

(1 minus 119905)3minus119894

119875119894 119899 = 3

(9)

26 Assessment Methodology The qualitative assessment andstatistical analysis is conducted to evaluate the performanceof traditional histogram equalization (referred to THE inSection 3) BBHE DSIHE RMSHE (119903 = 2) RSIHE (119903 = 2)and BBCCE Image visual quality is conducted by assess-ing the effect of improved tissue contrast on structuraldelineation relative to original image The radiologists are


1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)

Figure 2 Manifestations of contrast enhancement effect on MR image from medial central and lateral sides of the knee joint to representoverall enhancement impact on the stack of 2D MR images (First row from left to right) Original MR knee image from different sides (a)medial (b) central and (c) lateral (Second row from left to right) BBCCE enhanced MR knee image from different sides (d) medial (e)central and (f) lateral Knee cartilage is known as (in red arrows with label) (1) patellar cartilage (2) femoral cartilage (3) tibial cartilageProminent knee features in this image include (in white labeled arrows) (1) femoral sulcus and (2) Intensity variation within cartilage

consulted regarding the tissue contrast improvement effectproduced by different types of methods The performanceof various contrast enhancement methods from statisticalperspective in terms of contrast enhancement degree (EME)intensity distortion (AMBE) and image quality assessment(FSIM) is also examinedThe purpose of statistical analysis isto test the hypothesis that BBCCE outperforms fundamentalcontrast enhancement methods

Given an original image 119883 and its resultant image 119877measure of enhancement (EME) gauges the image enhance-ment of resultant image [22] Using a block size of 3 times 3 foreach of the nonoverlapping blocks 119887

11198872 EME estimates the

average contrast in every block by averaging the maximumintensity value 119868max and minimum intensity value 119868min Theimage contrast is enhanced if EME of image 119877 is greater thanimage 119883 In this experiment it is believed that the lesserEME values (but greater than original EME value) wouldbe adequate to improve tissue contrast in knee MR imageConsider the following

EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)

where Φ indicates orthogonal transforms and 120594(119909) is thesign function

Mean brightness difference between original image andresultant image reveals the degree of luminance distortionThe statistical evaluation metric known as absolute meanbrightness error (AMBE) is defined as

AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)

where119864(sdot) denotes the expectation119877 represents the resultantimage and 119883 represents the original image A lower AMBEis translated into better brightness preservation The AMBEare normalized and twenty MR images have been includedin experiment The formulation for average AMBE is shownin (12) For instance smaller AMBE implies lesser meanintensity distortion (lesser image information loss) AMBEvalue (AMBE = 0) is taken as reference from original MRimage

Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)

Feature similarity index model (FSIM) aims to performimage quality assessment (IQA) [23] for THE BBHEDSIHE


P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)

Figure 3 Central region manifestation of the left patellofemoral joint section using DESSwe MR imaging sequence Labels in (a) indicatevarious skeletal elements inside knee joint (P = patellar T = tibia F = femur 1 = femoral sulcus 2 = Hoffa fat pad 3 = anterior cruciateligament (ACL) 4 = oblique popliteus ligament 5 = posterior cruciate ligament (PCL) 6 = gastrocnemius muscle 7 = popliteus muscle8 = soleus muscle) For comparison purpose original MR image is remanifested in (b) Enhanced MR images using (c) THE (d) BBHE (e)DSIHE (f) RMSHE (g) RSIHE and (h) BBCCE show the contrast enhancement effect relative to (b)The femoral sulcus (white arrow labeledas ldquo1rdquo) and intensity variation within cartilage (unlabeled white arrow) are indicated in images from (b) to (h)

RMSHE RSIHE and BBCCE FSIM has the maximumscore of one indicating the highest enhanced image qualitySuppose that two images 119891

1and 119891

2are included to calculate

the similarity between these two images then FSIM is definedas follows

FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)

where PC119898(119909) is the dimensionless feature perceived at a

point where the Fourier components reachmaximal in phase119878119871(119909) is the overall similarity between images and 119891

1and 119891

2

andΩ are the whole image spatial domain

3 Results and Discussion

31 Qualitative Results We considered several factors whenassessing visual quality of BBCCE enhanced image such as


Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093


Table 2 The one-way ANOVA computed by using different contrast enhancement methods in EME AMBE and FSIM

Sum of squares df Mean square 119865 119875 valueEME

Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119

FSIMMethods 5846766 5 1169353 165239 000lowast

Errors 80675 114 708Total 5927440 119

AMBEMethods 042 5 009 12412 000lowast

Errors 008 114 000Total 050 119

lowastSignificant 119875 value (119875 lt 005)

Table 3 Categorization of different methods using Fisherrsquos Least Significance Difference (LSD) for EME

(119868) Method (119869) Method Mean difference (119868 minus 119869) Std error 119875 value 95 confidence intervalLower bound Upper bound

THE

BBHE 192lowast 056 000 081 304DSIHE 061 056 028 minus051 173RMSHE 429lowast 056 000 317 541RSIHE minus712lowast 056 000 minus823 minus600BBCCE 618lowast 056 000 507 730

BBHE

THE minus192lowast 056 000 minus304 minus081DSIHE minus131lowast 056 002 minus243 minus020RMSHE 237lowast 056 000 125 348RSIHE minus904lowast 056 000 minus1016 minus792BBCCE 426lowast 056 000 314 538

DSIHE

THE minus061 056 028 minus173 051BBHE 131lowast 056 002 020 243RMSHE 368lowast 056 000 256 480RSIHE minus773lowast 056 000 minus884 minus661BBCCE 557lowast 056 000 446 670

RMSHE

THE minus429lowast 056 000 minus541 minus317BBHE minus237lowast 056 000 minus348 minus125DSIHE minus368lowast 056 000 minus480 minus256RSIHE minus1141lowast 056 000 minus1253 minus1029BBCCE 189lowast 056 000 078 301

RSIHE

THE 712lowast 056 000 600 823BBHE 904lowast 056 000 792 1016DSIHE 773lowast 056 000 661 884RMSHE 1141lowast 056 000 1029 1253BBCCE 1330lowast 056 000 1218 1441

BBCCE

THE minus618lowast 056 000 minus730 minus507BBHE minus426lowast 056 000 minus538 minus314DSIHE minus557lowast 056 000 minus669 minus446RMSHE minus189lowast 056 000 minus301 minus078RSIHE minus1330lowast 056 000 minus1442 minus1218

lowastThemean difference is significant at the 005 level


Table 4 Categorization of contrast enhancement methods intohomogenous subset using the Duncan test for EME

Method 119873

Subset for alpha = 0051 2 3 4 5

BBCCE 20 4144RMSHE 20 4333BBHE 20 4570DSIHE 20 4701THE 20 4762RSIHE 20 5474Sig 100 100 100 028 100Means for groups in homogeneous subsets are displayedHarmonic Mean Sample Size = 20000

natural looking tissue contrast improvement preservationof knee features and minimum image artifact provocationEffect of BBCCE enhancement is illustrated in Figure 2As such BBCCE improves tissue contrast (red arrows withlabels (1) patellar cartilage (2) articular cartilage and (3)meniscal cartilage) without obliterating knee features (whitearrows with labels (1) femoral sulcus and (2) slight inten-sity variation within the cartilage) Thus cartilage becomesmore differentiable from its surrounding tissues and bonesBesides not much image noise has been amplified after con-trast enhancement Thus well defined cartilage delineationappropriate contrast enhancement preservation of pertinentcartilage features and minimal noise amplification producenatural appearance for the resultant MR image of knee

Besides BBCCE is compared to fundamental contrastenhancement techniques with relative to original MR imageKnee joint is a compound joint packed with various skele-tal elements Interpretation of complex MR image couldlead to human ambiguity as a result of unclear structuraldelineation and unpleasant background image illustrationFigure 3 demonstrates various knee components observedfrom MR image and the enhancement effect imposed bydifferent contrast enhancement methods

Typically all contrast enhancement methods improve thetissue contrast lifting the imagersquos background luminanceHowever apparent image noise amplification is detected insome previous contrast enhancement methods For exampleimage noise amplified by traditional histogram equalization(THE) is obvious especially at femur and tibia Irritatingimage artifact downgrades the visual quality Besides seriousnoise amplification is detected at popliteus muscle gastroc-nemius muscle and soleus muscle in images produced byRSIHE and RMSHE Thus THE RSIHE and RMSHE arelargely unsuitable for medical image contrast enhancement

Preservation of salient information describing knee car-tilage is imperative Precise structural delineation throughadequate tissue contrast refinement can maintain small butimportant feature details For instance BBCCE improves thecartilage contrast as well as conserves paramount details likefemoral sulcus (white arrow labeled as ldquo1rdquo) and intensityvariation within cartilage (unlabeled white arrow) simultane-ously Previous contrast enhancement methods on the other

hand over-enhance the cartilage Hence intensity variationis destroyed and patellar cartilage is seen to ldquocombinerdquo withfemoral cartilage

32 Statistical Analysis All tests are performed using SPSS(version 21) In this study the hypothesis that BBCCEcould outperform other contrast enhancement methods istested and verified Table 1 shows the mean values for EMEFSIM andAMBEof different contrast enhancementmethods(THE BBHE DSIHE RMSHE RSIHE BBCCE) using 20healthy subjects The results are evaluated by taking orig-inal image values (EME = 3228 AMBE = 0 FSIM = 0-1) as reference EME assesses the performance of contrastenhancement methods in terms of enhancement degree Itis deduced that BBCCE (4144 plusmn 106) produces the leastenhancement followed by RMSHE (4333 plusmn 180) AMBEassesses the performance of contrast enhancement methodsfrom intensity distortion perspective RMSHE (1402 plusmn 129)is ranked first in AMBE test and is followed closely by BBCCE(1482 plusmn 147) FSIM assesses the performance of contrastenhancementmethods fromquality of image perspective Forinstance BBCCE (092 plusmn 002) is ranked first in producinghighest FSIM score followed by RMSHE (083 plusmn 003) It isobserved that BBCCE and RMSHE produces almost similarresults and further analysis should be performed to test ourhypothesis

Table 2 indicates the contrast enhancement methodswhich impose significantly different impact on the MRimages in all cases that is EMEAMBE and FSIM (119875 lt 005)Therefore the data is further analyzed using post hoc tests(Fisherrsquos Least Significant Difference and the Duncan test) toevaluate the performance of different contrast enhancementmethods

Fisherrsquos least significant difference test (Table 3) indicatesthe mean differences between THE and DSIHE (061 minus061for DSIHE and THE) which is insignificant in EME Othercontrast enhancement methods show significant mean dif-ferenceThe result is confirmed by the Duncan test (Table 4)which categorized THE and DSIHE into the same subset

Fisherrsquos least significant difference test (Table 5) indicatesthe mean differences between RMSHE and BBCCE (minus079079 for BBCCE and RMSHE) which is insignificant inAMBE Other contrast enhancement methods show signifi-cant mean difference The result is confirmed by the Duncantest (Table 6) which categorized RMSHE and BBCCE in thesame subset

Fisherrsquos least significant difference test (Table 7) indicatesthe mean differences between RMSHE and RSIHE (minus001001 for RSIHE and RMSHE) which is insignificant in FSIMOther contrast enhancement methods show significant meandifference The result is confirmed by the Duncan test(Table 8) which categorized RMSHE and RSIHE in the samesubset

The performance of contrast enhancement methods(THE BBHE DSIHE RMSHE RSIHE and BBCCE) isranked according to the results computed from EME AMBEand FSIM in Table 9 Empty spaces that are observed fromTable 9 show the readjustment made after conducting post


Table 5 Categorization of different methods using Fisherrsquos Least Significance Difference (LSD) for AMBE


THE

BBHE 5004lowast 084 000 4838 5171DSIHE 4159lowast 084 000 3992 4325RMSHE 6337lowast 084 000 6170 6503RSIHE 6000lowast 084 000 5832 6166BBCCE 6257lowast 084 000 6091 6424

BBHE

THE minus5004lowast 084 000 minus5171 minus4838DSIHE minus845lowast 084 000 minus1012 minus679RMSHE 1333lowast 084 000 1166 1499RSIHE 995lowast 084 000 828 1162BBCCE 1253lowast 084 000 1086 1420

DSIHE

THE minus4159lowast 084 000 minus4325 minus3992BBHE 845lowast 084 000 679 1012RMSHE 2178lowast 084 000 2011 2345RSIHE 1840lowast 084 000 1674 2007BBCCE 2098lowast 084 000 1932 2265

RMSHE

THE minus6337lowast 084 000 minus6503 minus6170BBHE minus1333lowast 084 000 minus1500 minus1166DSIHE minus2178lowast 084 000 minus2345 minus2011RSIHE minus338lowast 084 000 minus504 minus171BBCCE minus079 084 035 minus246 087

RSIHE

THE minus6000lowast 084 000 minus6166 minus5832BBHE minus995lowast 084 000 minus1162 minus828DSIHE minus1840lowast 084 000 minus2007 minus1674RMSHE 338lowast 084 000 171 504BBCCE 258lowast 084 000 092 425

BBCCE

THE minus6257lowast 084 000 minus6424 minus6091BBHE minus1253lowast 084 000 minus1420 minus1087DSIHE minus2098lowast 084 000 minus2265 minus1932RMSHE 079 084 035 minus087 246RSIHE minus258lowast 084 000 minus425 minus092


Table 6 Categorization of contrast enhancement methods intohomogenous subset using the Duncan test for AMBE

Method 119873


RMSHE 20 1402BBCCE 20 1482RSIHE 20 1740BBHE 20 2735DSIHE 20 3580THE 20 7739Sig 035 100 100 100 100Means for groups in homogeneous subsets are displayedHarmonic Mean Sample Size = 20000

hoc tests For instance it is found that BBCCE is outper-formed consistently in all experiments BBCCE (bolded) isranked first in EME and FSIM while RMSHE is ranked

equally with BBCCE in AMBE Besides it is concluded thatTHE consistently occupies at near bottom of the rankingindicating the severe limitations faced by conventional his-togram equalization to improve medical image contrast

4 Conclusion

In this paper bihistogram Bezier curve contrast enhance-ment (BBCCE) is presented The objective of BBCCE is toassist radiologist to interpret MR image prior to performingknee cartilage segmentation However MR image possessespoor tissue contrast and conventional contrast enhancementmethods have failed to address this issue As such suddenjump in conventional transform curve causes the imageto be over-enhanced and deteriorates the image visualquality Therefore it is believed that the adequate contrastenhancement is the most appropriate solution as in the caseof MR image To achieve the objective the Bezier transformcurve based on intensity discrepancy value and intensity


Table 7 Categorization of different methods using Fisherrsquos Least Significance Difference (LSD) for FSIM


THE

BBHE minus007lowast 001 000 minus009 minus005DSIHE minus005lowast 001 000 minus006 minus003RMSHE minus011lowast 001 000 minus012 minus009RSIHE minus010lowast 001 000 minus011 minus008BBCCE minus019lowast 001 000 minus021 minus018

BBHE

THE 007lowast 001 000 005 009DSIHE 002lowast 001 001 000 004RMSHE minus004lowast 001 000 minus005 minus002RSIHE minus003lowast 001 000 minus004 minus001BBCCE minus012lowast 001 000 minus014 minus011

DSIHE

THE 005lowast 001 000 003 006BBHE minus002lowast 001 001 minus004 minus000RMSHE minus006lowast 001 000 minus008 minus004RSIHE minus005lowast 001 000 minus006 minus003BBCCE minus015lowast 001 000 minus016 minus013

RMSHE

THE 011lowast 001 000 009 012BBHE 003lowast 001 000 002 005DSIHE 006lowast 001 000 004 008RSIHE 001 001 012 minus000 003BBCCE minus009lowast 001 000 minus010 minus007

RSIHE

THE 009lowast 001 000 008 011BBHE 003lowast 001 000 001 004DSIHE 005lowast 001 000 003 006RMSHE minus001 001 012 minus003 000BBCCE minus010lowast 001 000 minus012 minus008

BBCCE

THE 019lowast 001 000 018 021BBHE 012lowast 001 000 011 014DSIHE 015lowast 001 000 013 016RMSHE 009lowast 001 000 007 010RSIHE 010lowast 001 000 008 012


Table 8 Categorization of contrast enhancement methods intohomogenous subset using Duncanrsquos test for FSIM

Method 119873


THE 20 072DSIHE 20 077BBHE 20 079RSIHE 20 082RMSHE 20 083BBCCE 20 092Sig 100 100 100 012 100Means for groups in homogeneous subsets are displayedHarmonic Mean Sample Size = 20000

difference curve is deduced The quantitative results showthat BBCCE excels in terms of tissue improvement minimalmean intensity distortion and image quality The results are

Table 9 Ranking of methods in terms of enhancement degree(EME) image quality (FSIM) and mean intensity distortion(AMBE) The methods ranking is computed according to FisherrsquosLeast Significance Difference (LSD) and the Duncan test

Rank EME FSIM AMBE1 BBCCE BBCCE RMSHE BBCCE2 RMSHE RMSHE RSIHE3 BBHE RSIHE4 THE DSIHE BBHE BBHE5 DSIHE DSIHE6 RSIHE THE THE

in-line with qualitative results which show that BBCCE couldpreserve important knee features and better delineate theknee structure without provoking much image noise In thefuture the mean opinion survey and record image evaluation


time to assess the effectiveness of BBCCE will be performedin assisting radiologists to interpret knee MR image

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors gratefully acknowledge the Ainuddin Wahidscholarship provided by School of Graduate Studies Uni-versiti Teknologi Malaysia and university research grantprovided by Research Management Centre and spon-sored by Ministry of Higher Education Malaysia VotQJ130000254504H41 GUP Universiti Teknologi MalaysiaJohor Bahru Malaysia

References

[1] F Eckstein C K Kwoh R M Boudreau et al ldquoQuantitativeMRI measures of cartilage predict knee replacement a case-control study from the Osteoarthritis Initiativerdquo Annals of theRheumatic Diseases vol 72 no 5 pp 707ndash714 2013

[2] F Eckstein D Burstein and T M Link ldquoQuantitative MRIof cartilage and bone degenerative changes in osteoarthritisrdquoNMR in Biomedicine vol 19 no 7 pp 822ndash854 2006

[3] J J Smith A G Sorensen and J H Thrall ldquoBiomarkers inimaging realizing radiologyrsquos futurerdquo Radiology vol 227 no 3pp 633ndash638 2003

[4] F Eckstein G Ateshian R Burgkart et al ldquoProposal for anomenclature for magnetic resonance imaging based measuresof articular cartilage in osteoarthritisrdquoOsteoarthritis and Carti-lage vol 14 no 10 pp 974ndash983 2006

[5] Y Boykov and G Funka-Lea ldquoGraph cuts and efficient N-Dimage segmentationrdquo International Journal of Computer Visionvol 70 no 2 pp 109ndash131 2006

[6] L Grady ldquoRandom walks for image segmentationrdquo IEEETransactions on Pattern Analysis and Machine Intelligence vol28 no 11 pp 1768ndash1783 2006

[7] E N Mortensen and W A Barrett ldquoInteractive segmentationwith intelligent scissorsrdquo Graphical Models and Image Process-ing vol 60 no 5 pp 349ndash384 1998

[8] W A Barrett and E N Mortensen ldquoInteractive live-wireboundary extractionrdquo Medical Image Analysis vol 1 no 4 pp331ndash341 1997

[9] Y Y Boykov andM-P Jolly ldquoInteractive graph cuts for optimalboundary amp region segmentation of objects in N-D imagesrdquoin Proceedings of the 8th International Conference on ComputerVision pp 105ndash112 July 2001

[10] Y Li J Sun C-K Tang and H-Y Shum Lazy Snapping ACMLos Angeles Calif USA 2004

[11] M Haubner F Eckstein M Schnier et al ldquoA non-invasivetechnique for 3-dimensional assessment of articular cartilagethickness based on MRI part 2 validation using CT arthrog-raphyrdquo Magnetic Resonance Imaging vol 15 no 7 pp 805ndash8131997

[12] R Gonzalez and R Woods Digital Image Processing PrenticeHall New York NY USA 2nd edition 2002

[13] W ZW Ismail and K S Sim ldquoContrast enhancement dynamichistogram equalization for medical image processing applica-tionrdquo International Journal of Imaging Systems and Technologyvol 21 no 3 pp 280ndash289 2011

[14] H Chai T Swee G Seng and L Wee ldquoMultipurpose contrastenhancement on epiphyseal plates and ossification centers forbone age assessmentrdquo BioMedical Engineering OnLine vol 12no 1 article 27 pp 1ndash19 2013

[15] Y-T Kim ldquoContrast enhancement using brightness preservingbi-histogram equalizationrdquo IEEE Transactions on ConsumerElectronics vol 43 no 1 pp 1ndash8 1997

[16] YWangQ Chen and B Zhang ldquoImage enhancement based onequal area dualistic sub-image histogram equalizationmethodrdquoIEEE Transactions on Consumer Electronics vol 45 no 1 pp68ndash75 1999

[17] S-D Chen and A R Ramli ldquoMinimum mean brightnesserror bi-histogramequalization in contrast enhancementrdquo IEEETransactions on Consumer Electronics vol 49 no 4 pp 1310ndash1319 2003

[18] S-D Chen and A R Ramli ldquoContrast enhancement usingrecursive mean-separate histogram equalization for scalablebrightness preservationrdquo IEEE Transactions on Consumer Elec-tronics vol 49 no 4 pp 1301ndash1309 2003

[19] K S Sim C P Tso and Y Y Tan ldquoRecursive sub-imagehistogram equalization applied to gray scale imagesrdquo PatternRecognition Letters vol 28 no 10 pp 1209ndash1221 2007

[20] P A Hardy M P Recht D Piraino and D ThomassonldquoOptimization of a dual echo in the steady state (DESS) free-precession sequence for imaging cartilagerdquo Journal of MagneticResonance Imaging vol 6 no 2 pp 329ndash335 1996

[21] R H Bartels J C Beatty and B A Barsky An Introduction toSplines for Use in Computer Graphics and Geometric ModelingMorgan Kaufmann Boston Mass USA 1987

[22] S S Agaian K Panetta and A M Grigoryan ldquoTransform-based image enhancement algorithms with performance mea-surerdquo IEEE Transactions on Image Processing vol 10 no 3 pp367ndash382 2001

[23] L Zhang L Zhang X Mou and D Zhang ldquoFSIM a featuresimilarity index for image quality assessmentrdquo IEEE Transac-tions on Image Processing vol 20 no 8 pp 2378ndash2386 2011

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



1 Introduction

Magnetic resonance (MR) imaging allows direct visualizationof knee cartilage and quantitative measurement on cartilageto monitor osteoarthritis (OA) progression [1] The advance-ment of MR imaging contributes greatly to the search foreffective OA imaging biomarker [2] Imaging biomarker isdefined as ldquoany anatomic physiologic biochemical ormolec-ular parameter detectable with one ormore imagingmethodsused to diagnose the presence andor severity of disease [3]rdquoSince imaging biomarker is crucial for developing diseasemodifying osteoarthritis drug (DMOAD) [4] the NationalInstitutions of Health (NIH) has launched the OsteoarthritisInitiative (OAI) in a public-private consortium To date atotal of 4796 participants have been recruited from four clini-cal institutions (University of Maryland School of MedicineBaltimore Md Ohio State University Columbus OH Uni-versity of Pittsburgh Pittsburgh Pa Memorial Hospital ofRhode Island Pawtucket RI) to provide comprehensiveanalysis for OA progression and identification of suitable OAbiomarkers

There have been growing interests in interactive segmen-tation recently [5ndash8]Medical images such as kneeMR imagesare usually too complex for fully automated segmentation toproduce satisfactory results On the other hand interactivesegmentation reflects user intention because the segmenta-tion is initiated according to minimal user intervention inthe forms of scribbles [9] or bounding box [10] Thereforeunnecessary segmentation error such as oversegmentationreported in automated segmentation [11] can be avoidedTypically interactive segmentation can be divided into twogroups boundary based and contour based Although bothgroups present different segmentation approaches initializa-tion of these methods will greatly depend on human knowl-edge Therefore direct incorporation of human knowledgeto commence the segmentation translates into underlyingneed for human interpretation of MR image Distinct featuremanifestation and excellent visual quality of the medicalimage for instance play important role to assist user ininterpreting medical images and then inserting seeds ontoregion of interests

The visual quality of knee MR image can be deterioratedby several phenomena insignificant tissue contrast com-plex knee joint structure and low background luminanceDefinitive tissue contrast facilitates the image interpretationprocess Unfortunately mediocre tissue contrast has beenobserved in knee MR image For example tissue contrastbetween cartilage surrounding muscle tissues and imagebackground only differ slightly As a result poor contrastdifference observed in the image can easily lead to inter- andintraobserver ambiguity during inspection Besides humanknee structure is anatomically complex The knee compart-ment is filled with various types of cartilages knee bonesmuscle fat tissue and ligaments attempting to interpretMR image equipped with poor tissue contrast is laboriousFurther MR image of knee is characterized by large numbersof low intensity background pixels Dark background hard-ens any effective identification of pertinent image featuresEventually above mentioned phenomena give rise to the

prominence of a tissue contrast enhancement model toelevate image visual quality

Contrast enhancement is defined as a remapping processto transform the imagersquos intensity distribution so the imageintensity range can be fully exploited [12] In particularcontrast enhancement has been implemented to extractsalient information protrude image features and improvethe imagersquos visual quality Applications of various contrastenhancement models on medical images for further analysishave been reported in previous studies Ismail and Sim [13]applied dynamic contrast enhancement on low resolutionbrain MR images so important brain image features couldbe protruded Chai et al [14] proposed the use of multipur-pose contrast enhancement technique to overcome the poorcontrast problem reported in the ossification sites of childrenhand bones In the context of knee MR image adequatecontrast enhancement to ameliorate the image appearancewill be meaningful enough for the subsequent interactiveknee cartilage segmentation to succeed

Direct contrast enhancement through traditional his-togramequalization [12] cannot be implemented intomedicalimage processing Traditional histogram equalization distortsthe brightness enhancement by invariably shifting outputmean brightness to middle gray level The resultant imagetherefore is over-enhanced and any brightness sensitiveimage feature is destroyed Besides traditional histogramequalization combines any neighboring gray levels with lightprobabilistic density into one gray level widening the gapwith neighboring gray levels which have heavy probabilisticdensity The flawed intensity redistribution contributes tosevere detail image loss and sudden jump in cumulativedensity function

Fundamental contrast enhancement betterments aredeveloped based on the idea of histogram partitioning Thegist of these improved models concentrates on preservingmean brightness of original image using different intensitythreshold values that is mean and median As such oneof the earliest bihistogram equalization methods brightnesspreserving bihistogram equalization (BBHE) [15] has beendeveloped based on this concept BBHE partitions the globalhistogram into two subhistograms according to mean inten-sity value Then traditional histogram equalization is imple-mented independently onto the subhistograms Given thatBBHE partitions the intensity distribution into two whilemean brightness of input image is preserved Conservation ofmean brightness is deemed imperative in preventing resultantimage from severe intensity distortion

In 1999 Wang et al [16] introduced dualistic subimagehistogram equalization (DSIHE) This method uses medianintensity value to decompose original imagersquos histogramYu postulates that maximum segmentation entropy can beobtained when two subimages have equal areas DSIHEis capable of preserving original imagersquos mean brightnessas well as attaining the most image entropy Meanwhileminimum mean brightness error bihistogram equalization(MMBEBHE) [17] has developed based on iterative selec-tion of optimal threshold value For instance the optimalthreshold valuemust satisfy minimum absolute mean bright-ness error (AMBE) between output histogram and input












119883119900 1198831 119883







119901 (119883119896) =

119899119896

119899

(1)


120583 =

sum119896

119896=0119899119896(119883119896)

119899

=

119896

sum

119896=0

119901 (119883119896)119883119896 (2)


119896against 119883

119896 To attain



(a)

0 50 100 150 200 250 3000

1000

2000

3000

4000 PDF

Intensity

Num

ber o

f pix

els

0

1000

2000

3000

4000 Lower PDF

Num

ber o

f pix

els

0 50 100 150 200 250 3000

500

1000

1500 Upper PDF

Intensity

Num

ber o

f pix

el

0 50 100 150 200 250 300Intensity

(b)

0 50 100 150 200 250 3000

50

100

150

200

250

300 CDF

Intensity of pixel

Cum

ulat

ive f

requ

ency

of p

ixels

(c)

0

k

l

50 100 150 200 250 300

0

20

40

60

80

100

120

L

(59 0)

(44 5)

(125 107)

(17 minus11)minus20

ΔL

(d)

0 50 100 150 200 250 3000

50

100

150

200

250

300

Original intensity

Enha

nced

inte

nsity

(e) (f)




119888 (119909) =

119896

sum

119896=119900

119901 (119883119896)

119891 (119909) = 119883119900+ (119883119871minus1

minus 119883119900) 119888 (119909)

(3)



AID = 1199091015840

minus 119909 (4)




0 otherwise


0 otherwise

(5)




0 1198751 119875



119876 (119905) =

119899

sum

119894=0

119861119899119894(119905) 119875119894 (6)



119861119899119894(119905) = (

119899

119894) 119905119894

(1 minus 119905)119899minus119894

(7)


(

119899

119894) =

119899

119894 (119899 minus 1)

if 0 le 119894 le 119899

0 otherwise(8)





119876 (119905) =

2

sum

119894=0

(

2

119894

) 119905119894

(1 minus 119905)2minus119894

119875119894 119899 = 2

3

sum

119894=0

(

3

119894

) 119905119894

(1 minus 119905)3minus119894

119875119894 119899 = 3

(9)



1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)






EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)



AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)


Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)



P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in














119883119900 1198831 119883







119901 (119883119896) =

119899119896

119899

(1)


120583 =

sum119896

119896=0119899119896(119883119896)

119899

=

119896

sum

119896=0

119901 (119883119896)119883119896 (2)


119896against 119883

119896 To attain



(a)

0 50 100 150 200 250 3000

1000

2000

3000

4000 PDF

Intensity

Num

ber o

f pix

els

0

1000

2000

3000

4000 Lower PDF

Num

ber o

f pix

els

0 50 100 150 200 250 3000

500

1000

1500 Upper PDF

Intensity

Num

ber o

f pix

el

0 50 100 150 200 250 300Intensity

(b)

0 50 100 150 200 250 3000

50

100

150

200

250

300 CDF

Intensity of pixel

Cum

ulat

ive f

requ

ency

of p

ixels

(c)

0

k

l

50 100 150 200 250 300

0

20

40

60

80

100

120

L

(59 0)

(44 5)

(125 107)

(17 minus11)minus20

ΔL

(d)

0 50 100 150 200 250 3000

50

100

150

200

250

300

Original intensity

Enha

nced

inte

nsity

(e) (f)




119888 (119909) =

119896

sum

119896=119900

119901 (119883119896)

119891 (119909) = 119883119900+ (119883119871minus1

minus 119883119900) 119888 (119909)

(3)



AID = 1199091015840

minus 119909 (4)




0 otherwise


0 otherwise

(5)




0 1198751 119875



119876 (119905) =

119899

sum

119894=0

119861119899119894(119905) 119875119894 (6)



119861119899119894(119905) = (

119899

119894) 119905119894

(1 minus 119905)119899minus119894

(7)


(

119899

119894) =

119899

119894 (119899 minus 1)

if 0 le 119894 le 119899

0 otherwise(8)





119876 (119905) =

2

sum

119894=0

(

2

119894

) 119905119894

(1 minus 119905)2minus119894

119875119894 119899 = 2

3

sum

119894=0

(

3

119894

) 119905119894

(1 minus 119905)3minus119894

119875119894 119899 = 3

(9)



1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)






EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)



AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)


Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)



P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




(a)

0 50 100 150 200 250 3000

1000

2000

3000

4000 PDF

Intensity

Num

ber o

f pix

els

0

1000

2000

3000

4000 Lower PDF

Num

ber o

f pix

els

0 50 100 150 200 250 3000

500

1000

1500 Upper PDF

Intensity

Num

ber o

f pix

el

0 50 100 150 200 250 300Intensity

(b)

0 50 100 150 200 250 3000

50

100

150

200

250

300 CDF

Intensity of pixel

Cum

ulat

ive f

requ

ency

of p

ixels

(c)

0

k

l

50 100 150 200 250 300

0

20

40

60

80

100

120

L

(59 0)

(44 5)

(125 107)

(17 minus11)minus20

ΔL

(d)

0 50 100 150 200 250 3000

50

100

150

200

250

300

Original intensity

Enha

nced

inte

nsity

(e) (f)




119888 (119909) =

119896

sum

119896=119900

119901 (119883119896)

119891 (119909) = 119883119900+ (119883119871minus1

minus 119883119900) 119888 (119909)

(3)



AID = 1199091015840

minus 119909 (4)




0 otherwise


0 otherwise

(5)




0 1198751 119875



119876 (119905) =

119899

sum

119894=0

119861119899119894(119905) 119875119894 (6)



119861119899119894(119905) = (

119899

119894) 119905119894

(1 minus 119905)119899minus119894

(7)


(

119899

119894) =

119899

119894 (119899 minus 1)

if 0 le 119894 le 119899

0 otherwise(8)





119876 (119905) =

2

sum

119894=0

(

2

119894

) 119905119894

(1 minus 119905)2minus119894

119875119894 119899 = 2

3

sum

119894=0

(

3

119894

) 119905119894

(1 minus 119905)3minus119894

119875119894 119899 = 3

(9)



1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)






EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)



AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)


Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)



P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





119888 (119909) =

119896

sum

119896=119900

119901 (119883119896)

119891 (119909) = 119883119900+ (119883119871minus1

minus 119883119900) 119888 (119909)

(3)



AID = 1199091015840

minus 119909 (4)




0 otherwise


0 otherwise

(5)




0 1198751 119875



119876 (119905) =

119899

sum

119894=0

119861119899119894(119905) 119875119894 (6)



119861119899119894(119905) = (

119899

119894) 119905119894

(1 minus 119905)119899minus119894

(7)


(

119899

119894) =

119899

119894 (119899 minus 1)

if 0 le 119894 le 119899

0 otherwise(8)





119876 (119905) =

2

sum

119894=0

(

2

119894

) 119905119894

(1 minus 119905)2minus119894

119875119894 119899 = 2

3

sum

119894=0

(

3

119894

) 119905119894

(1 minus 119905)3minus119894

119875119894 119899 = 3

(9)



1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)






EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)



AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)


Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)



P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




1

2

3

1

2

(a)

1

2

(b)

2

(c)

1

2

(d)

1

2

(e)

2

(f)






EME = maxΦisinΦ

120594(

1

11988711198872

1198872

sum

119897=1

1198871

sum

119896=1

20 log119868max119868min

) (10)



AMBE (119877119883) = |119864 (119877) minus 119864 (119883)| (11)


Average AMBE = 1

119873

119873

sum

119894=1

AMBE (119877119894 119883119894) (12)



P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




P F

T

1

2

3 4

56

7

8

(a)

1

(b)

1

(c)

1

(d)

1

(e)

1

(f)

1

(g)

1

(h)



1and 119891



FSIM =

sum119883isinΩ

119878119871(119909) sdot PC

119898(119909)

sum119883isinΩ

PC119898(119909)

(13)



1and 119891

2





Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Table1MeanEM

EAMBE

and

FSIM

values

compu

tedfro

mTH

EBB

HE

DSIHE

RMSH

ERS

IHE

andBB

CCEby

taking

originalim

agersquos

EME(3238)a

ndAMBE

(000)

asreference

FSIM

ranged

from

0to

1where

1ind

icates

theb

estenh

ancedim

ageq

uality

Metho

dsMeanEM

EplusmnSD

95confi

denceintervalof

thed

ifference

MeanAMBE

plusmnSD

95confi

denceintervalof

thed

ifference

MeanFSIM

plusmnSD

95confi

denceintervalof

thed

ifference

Lower

Upp

erLo

wer

Upp

erLo

wer

Upp

erTH

E4762plusmn205

4666

4858

7739plusmn499

7505

7973

072plusmn002

071

074

BBHE

4570plusmn192

4479

4660

2735plusmn261

2613

2857

079plusmn003

078

081

DSIHE

4701plusmn181

4616

4786

3580plusmn234

3471

3690

077plusmn003

076

079

RMSH

E4333plusmn180

4248

4417

140plusmn129

1342

1463

083plusmn003

082

085

RSIH

E5474plusmn188

5386

5562

1740plusmn119

1684

1796

082plusmn003

081

083

BBCC

E4144plusmn106

4094

4193

1482plusmn147

1413

1551

092plusmn002

091

093




Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






Methods 211180 5 42236 13272 000lowast

Errors 36280 114 318Total 247460 119


Errors 80675 114 708Total 5927440 119


Errors 008 114 000Total 050 119




THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Method 119873

















THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873





4 Conclusion





THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






THE


BBHE


DSIHE


RMSHE


RSIHE


BBCCE




Method 119873











Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in







Acknowledgments


References































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in










Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



research article medical image visual appearance improvement using...

Documents