Research ArticleAppropriate Contrast Enhancement Measures forBrain and Breast Cancer Images

Suneet Gupta1 and Rabins Porwal2

1Research Scholar Mewar University Chittorgarh 312901 Rajasthan India2Department of Information Technology Institute of Technology amp Science (ITS)Ghaziabad 201007 Uttar Pradesh India

Correspondence should be addressed to Suneet Gupta suneetsonuyahoocom

Received 12 November 2015 Accepted 15 March 2016

Academic Editor Richard H Bayford

Copyright copy 2016 S Gupta and R Porwal This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Medical imaging systems often produce images that require enhancement such as improving the image contrast as they are poorin contrast Therefore they must be enhanced before they are examined by medical professionals This is necessary for properdiagnosis and subsequent treatment We do have various enhancement algorithms which enhance the medical images to differentextents We also have various quantitative metrics or measures which evaluate the quality of an imageThis paper suggests the mostappropriate measures for two of the medical images namely brain cancer images and breast cancer images

1 Introduction

We have different methods of enhancements which enhancethe quality of an image [1 2] Enhancement can be done insuch a way that the quality of image is upgraded visuallyor suiting to an application [3 4] Example of the lattercase is medical image where the enhanced image must bemore informative rather than being visually more pleasingTherefore to assess the quality of image we have differentmeasures of enhancement There are some of the measureswhich suite images having different types of contents andattributes Contrast based measures of enhancement meth-ods especially in spatial domain use the intensities of pixelsfor calculation purpose These spatial contrast measures areusually derived fromWeber-Fechner law Michelson contrastmeasure [5] or Contrast Ratio (CR) The measures PSNREME AMEE [6] SOE EC and so forth are examples ofsuch spatial contrast measures [7 8] These measures highlydepend on image attributes like image content background(uniformnonuniform) lighting texture patterns which areperiodic single or multiple targets randomness and alsonoise and distortions Some measures or metrics are bestsuited to a set of attributes while other measures are bestsuited to other set of attributes If a measure cannot properly

tackle a particular image property the metric or measure willnot turn to prove a good measure for that image Also bylooking at the image attributes suitability of some measurescan be suggested In this paper we will suggest the mostappropriate or suitable measures of enhancement for brainand breast cancer images

2 Basic Measures of Image Contrast

21 Michelson Contrast This measure is generally used forimages which have equivalent bright and dark features forexample periodic patterns such as a sinusoidal gratingMichelson contrast is defined as

119862119872

=119868max minus 119868min119868max + 119868min

(1)

where 119868max and 119868min are the highest and lowest intensities

22 Weber-Fechner Law This measure is generally usedfor images with a large uniform background and a smalltest target It is used for images whose average intensity

Hindawi Publishing CorporationInternational Journal of Biomedical ImagingVolume 2016 Article ID 4710842 8 pageshttpdxdoiorg10115520164710842

2 International Journal of Biomedical Imaging

is approximately equal to background intensity It is givenas

119862119882

=119868119905minus 119868119887

119868119887

(2)

where 119868119905is the target intensity and 119868

119887is the intensity of the

immediate adjacent background

23 Contrast Ratio Thismeasure is expressed in either linearor logarithmic form

CR =119868119905

119868119887

log (CR) =log (119868119905)

log (119868119887)

(3)

where 119868119905is the target intensity and 119868

119887is the intensity of the

immediate adjacent background

24 Entropy Entropy is calculated over entire image fromthe histogram of that image It is a measure of randomnesscharacterizing the texture of the image

Entropy = minussum119901 times ln (119901) (4)

where 119901 is the histogram count for a segment of imageIncreasing the contrast in one portion of an image anddecreasing it in another portionmay result in similar entropyas the original imageThis is because the entropy is calculatedover the entire image

3 Conventional Measures of Contrast

These measures are attractive because they are simple andeasy to calculateMSE andPSNRconsider corresponding pairof pixels of two images for calculation purpose The conven-tional measures actually calculate the difference between twoimages and therefore it can be used for both restoration andenhancement These measures are taken from [5]

31 Mean Squared Error (MSE) Let 119860(119898 119899) and 119861(119898 119899) bethe original and enhanced image respectively where 119898 times 119899

represents the size of image Then MSE is defined as

MSE =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119895=0

(119861 (119894 119895) minus 119860 (119894 119895))2 (5)

32 Peak Signal-to-Noise Ratio (PSNR) Let 119860(119898 119899) be animagewith119871 gray levels and119898times 119899 being the sizeThen PSNRis defined as

PSNR = 10 log10((119871 minus 1)

2

MSE) (6)

33 AbsoluteMeanBrightness Error (AMBE) Let119860(119898 119899) and119861(119898 119899) be the original and enhanced image respectively

where119898times119899 represents the size of image AMBE is calculatedfrom the following equation

AMBE = |119864 [119884] minus 119864 [119883]| (7)

where119864[119884] and119864[119883] aremean gray levels of new and originalimage respectively They are given as

119864 [119883] =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119894=0

119860 (119894 119895)

119864 [119884] =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119894=0

119861 (119894 119895)

(8)

4 Complex Measures of Contrast

These measures are based on the basic contrast measuresand are highly sensitive to randomness periodicity uniformbackground texture target size and noise These measuresare taken from [6]These complexmeasures are always a goodchoice as goodmeasures and by looking at the attributes oneor two measures can be suggested But doing it quantitativelyis always a better option Figure 1 [6 9ndash11] shows imageshaving different attributes

Let 119860(119898 119899) be an image which is split into 119903 times 119888 blockswith 119898 times 119899 being the size of the image Let each block berepresented as 119861

119896119897(119894 119895) where 119894 and 119895 refer to 119894th row and 119895th

column respectively inside each block Let 119868max119896119897 and 119868min119896119897be the maximum andminimum intensities in an image block119861119896119897(119894 119895)

41 Measure of Enhancement (EME) The EME is defined as

EME119903119888=

1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln(119868max119896119897

119868min119896119897)] (9)

As defined in (3) the CR of a block 119861119896119897is

CR119896119897

=119868max119896119897

119868min119896119897 (10)

Therefore

EME119903119888=

1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (CR119896119897)] (11)

The EME measure is suitable for images with attributes likenoncomplex segments uniform background in segmentssmall targets in segments nonperiodic pattern in segmentsand little to no randomness in segments

42Measure of Enhancement by Entropy (EMEE) TheEMEEis defined as

EMEE120572119903119888

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897

119868min119896119897)

120572

ln(119868max119896119897

119868min119896119897)]

(12)

International Journal of Biomedical Imaging 3

(a) Large uniform background small targetsno randomness and no periodic patterns

(b) Semiperiodic nonuniform back-ground

(c) Large uniform background nonperi-odic pattern

(d) Periodic nonuniform background (e) Random texture nonperiodic nonuni-form background

(f) Random texture nonperiodic nonuni-form background

(g) Nonperiodic nonuniform back-ground

(h) Highly textured nonperiodic nonuniform back-ground

(i) Little texture nonperiodic non-uniform background

Figure 1 Images with different attributes

As defined in (3) the CR of a block 119861119896119897is

CR119896119897

=119868max119896119897

119868min119896119897 (13)

Therefore

EMEE120572119903119888

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (CR119896119897)120572 ln (CR

119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (CR120572)]

(14)

Therefore EMEEmeasure is the entropy of Contrast Ratio foreach block 119861

119896119897 scaled by 120572 averaged over the entire image

The EMEE measure is suitable for images with attributeslike noncomplex segments or small size complex segmentsnonperiodic patterns in segments and randomness because

4 International Journal of Biomedical Imaging

of the inclusion of entropyThemore the value of 120572 the morethe emphasis on entropy The term 120572 helps to handle morerandomness

43 Logarithmic Michelson Contrast Measure (AME) TheAME is defined as

AME119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (119862119872119896119897

)] (15)

The AME measure is suitable for images with attributes likeperiodic patterns in segments and no randomness in textureIt cannot be used for images with uniform background andalso images with segments of large uniform intensity

44 Logarithmic AME by Entropy (AMEE) The AMEE isdefined as

AMEE119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)

120572

sdot ln(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)]

(16)

Michelson contrast is given as

119862119872

=119868max minus 119868min119868max + 119868min

(17)

Therefore

AMEE120572119903119888

= minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (119862119872119896119897

)120572 ln (119862

119872119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (119862119872119896119897

)120572]

(18)

Therefore AMEE measure is the entropy of Michelsoncontrast for each block 119861

119896119897 scaled by 120572 averaged over the

entire imageThe AMEE measure is suitable for images with the

attributes like surplus randomness in texture and periodicpatterns in segments It cannot handle images with largeuniform background

5 Other Measures of Contrast

Here the first two measures are taken from [7] and the nexttwo from [8]

51 Discrete Gray Level Energy (GLE) The gray level energymeasure basically refers to the way of distribution of graylevels It is given as

GLE =

119896

sum

119894=1

119901 (119894)2 (19)

where GLE represents the gray level energy with 256 binsand 119901(119894) refers to the probability distribution functions of

different gray levels that is the histogram count In specialcase GLE reaches its maximum value of 1 when the imagehas a constant value of gray level

GLEmax = max119896

sum

119894=1

119901 (119894)2 = 1 (20)

52 Relative Entropy (RE) Let 119901 and 119902 be two probabilityfunctions of processing images then RE of 119901 with respectto 119902 is defined as

RE =

119896

sum

119894=1

119901 (119894) log2

119901 (119894)

119902 (119894) (21)

Relative entropy is also called the Kullback Leibler distance

53 Second-Order Entropy (SOE) The first-order entropy isused as enhancement metric as it is less sensitive to noiseand has greater response to gray level step But it has thedrawback of ignoring the correlation among pixels Second-order derivatives enhance noise points much more and theyare more sensitive to a line than a step and to a point than toa line

The SOE as defined in [12] uses a cooccurrence matrix toconsider the spatial correlation Let119860 be an image of size 119909 times

119910with 119871 gray levels then element in the cooccurrencematrix119862 of size 119871 times 119871 which contains the information of correlationof the image is given as

119888119894119895

=

119910minus1

sum

119897=0

119909minus1

sum

119896=0

120595 (119897 119896) (22)

where 119888119894119895

is element at 119894th row and 119895th column of matrix 119862

and

120595 (119897 119896) = 1 if (

119860(119897 119896) = 119894 119860 (119897 119896 + 1) = 119895

andor119860 (119897 119896) = 119894 119860 (119897 + 1 119896) = 119895

)

120595 (119897 119896) = 0 otherwise

(23)

The probability of cooccurrence 119901119894119895is computed as

119901119894119895

=119888119894119895

sum119871minus1

119896=0sum119871minus1

119897=0119888119894119895

(24)

and the SOE is computed as

SOE = minussum

119895

sum

119894

119901119894119895log2(119901119894119895) (25)

54 Edge Content (EC) Let 119860(119909 119910) be an image where 119909

is the row coordinate and 119910 is the column coordinate Thegradient vector at any pixel position (119909 119910) is computed as

120575119860 (119909 119910) = [

119866119909

119866119910

] =

[[[[

[

120597

120597119909119860 (119909 119910)

120597

120597119910119860 (119909 119910)

]]]]

]

(26)

International Journal of Biomedical Imaging 5

Table 1 Average metrics values for brain cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 85517 265507 663417 1097868 1845359PSNR mdash 388112 338924 299151 277284 254730AMBE mdash 26961 47250 74828 95790 124371EME 40391 43321 45948 49485 52719 56941EMEE 00219 00236 00251 00272 00292 00317AME 698246 686713 678564 665659 656326 642347AMEE 02166 02138 02118 02087 02063 02029GLE 05199 05226 05265 05289 05331 05363RE mdash 128020 129135 130003 131236 132628SOE 37928 37594 37108 36806 36304 35931EC 09306 09801 10253 10805 11305 11884

where

119866119909= 119860 (119909 119910) minus 119860 (119909 + 1 119910)

119866119910= 119860 (119909 119910) minus 119860 (119909 119910 + 1)

(27)

Taking absolute value of image gradient as an indicator ofcontrast

|120575119860| = radic1198662119909+ 1198662119910 (28)

The measure EC is defined as

EC =1

119903 times 119888sum

119909

sum

119910

1003816100381610038161003816120575119860 (119909 119910)1003816100381610038161003816 (29)

where 119903 times 119888 is the size of image and

1 le 119909 le 119903

1 le 119910 le 119888

(30)

Contrast changes exist over the entire image and EC takesinto account all the contrast changes even for pixels whichare very close (adjacent) to each other

6 Enhancement Procedure

To decide the most appropriate contrast measure anenhancement procedure is required which enhances theimage contrast to different levels For this purpose a Matlabfunction ldquoimadjust()rdquo [13] is used to increase the contrastto different levels First an image is read and if required

converted to gray image Let us call this image the origi-nal image From the original image an adjusted image ofreasonably poor contrast is obtained Let us call this imagereference image The reference image is created using theMatlab command as follows

Reference Image RI = imadjust(Original Image[02 08] [02 08])

where the second parameter that is the input range of graylevels gets mapped to the third parameter that is the outputrange of gray levelsThe full range of gray levels [0 1] for bothinput and output refers to [0 255]

Now using the reference image RI different images ofdifferent levels of enhancement are obtained as follows

1198641 = imadjust(RI [021 079] [019 081])1198642 = imadjust(RI [022 078] [018 082])1198643 = imadjust(RI [023 077] [017 083])1198644 = imadjust(RI [024 076] [016 084])1198645 = imadjust(RI [025 075] [015 085])

where 1198641 1198642 1198643 1198644 and1198645 are five enhanced versions ofthe reference image RIThe enhancements of images are suchthat the contrast of RI lt 1198641 lt 1198642 lt 1198643 lt 1198644 lt 1198645 It canbe seen that for every next level of enhancement the inputrange has been narrowed down and the output range has beenexpanded resulting in contrast stretching and therefore morecontrast

7 Tests and Results

For the purpose of testing twenty medical images are takenfrom [14] ten brain cancer images and ten breast cancerimages From each of these images six versions namely onereference image (RI) and five enhanced images (1198641 to 1198645)are created as discussed above One example of each categorynamely brain cancer and breast cancer is shown in Figures 2and 3 Then the average values of the contrast enhancementmeasures or metrics as mentioned above are computed andthe results are shown in Tables 1 and 2 As far as the complexmeasures EME EMEE AME and AMEE are concerned theimages are divided into blocks of size 16 times 16 and in caseof EMEE and AMEE the value for 120572 is taken as 01 as therandomness in the images of brain and breast cancer is quitelow

Defining contrast stretching (CS) quantitatively withrespect to the reference image (RI) for the different enhancedversions of the images is as given in the following

CS =((HighOut minus LowOut) (HighIn minus LowIn)) of Enhanced Image((HighOut minus LowOut) (HighIn minus LowIn)) of Reference Image

(31)

where [LowOut HighOut] and [LowIn HighIn] are outputrange and input range of gray levels of the image Thedenominator of (31) is 1 since

HighOut minus LowOut

HighIn minus LowInof Reference Image = 08 minus 02

08 minus 02

= 10

(32)

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

2 International Journal of Biomedical Imaging

is approximately equal to background intensity It is givenas

119862119882

=119868119905minus 119868119887

119868119887

(2)

where 119868119905is the target intensity and 119868

119887is the intensity of the

immediate adjacent background

23 Contrast Ratio Thismeasure is expressed in either linearor logarithmic form

CR =119868119905

119868119887

log (CR) =log (119868119905)

log (119868119887)

(3)

where 119868119905is the target intensity and 119868

119887is the intensity of the

immediate adjacent background

24 Entropy Entropy is calculated over entire image fromthe histogram of that image It is a measure of randomnesscharacterizing the texture of the image

Entropy = minussum119901 times ln (119901) (4)

where 119901 is the histogram count for a segment of imageIncreasing the contrast in one portion of an image anddecreasing it in another portionmay result in similar entropyas the original imageThis is because the entropy is calculatedover the entire image

3 Conventional Measures of Contrast

These measures are attractive because they are simple andeasy to calculateMSE andPSNRconsider corresponding pairof pixels of two images for calculation purpose The conven-tional measures actually calculate the difference between twoimages and therefore it can be used for both restoration andenhancement These measures are taken from [5]

31 Mean Squared Error (MSE) Let 119860(119898 119899) and 119861(119898 119899) bethe original and enhanced image respectively where 119898 times 119899

represents the size of image Then MSE is defined as

MSE =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119895=0

(119861 (119894 119895) minus 119860 (119894 119895))2 (5)

32 Peak Signal-to-Noise Ratio (PSNR) Let 119860(119898 119899) be animagewith119871 gray levels and119898times 119899 being the sizeThen PSNRis defined as

PSNR = 10 log10((119871 minus 1)

2

MSE) (6)

33 AbsoluteMeanBrightness Error (AMBE) Let119860(119898 119899) and119861(119898 119899) be the original and enhanced image respectively

where119898times119899 represents the size of image AMBE is calculatedfrom the following equation

AMBE = |119864 [119884] minus 119864 [119883]| (7)

where119864[119884] and119864[119883] aremean gray levels of new and originalimage respectively They are given as

119864 [119883] =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119894=0

119860 (119894 119895)

119864 [119884] =1

119898 times 119899

119898minus1

sum

119894=0

119899minus1

sum

119894=0

119861 (119894 119895)

(8)

4 Complex Measures of Contrast

These measures are based on the basic contrast measuresand are highly sensitive to randomness periodicity uniformbackground texture target size and noise These measuresare taken from [6]These complexmeasures are always a goodchoice as goodmeasures and by looking at the attributes oneor two measures can be suggested But doing it quantitativelyis always a better option Figure 1 [6 9ndash11] shows imageshaving different attributes

Let 119860(119898 119899) be an image which is split into 119903 times 119888 blockswith 119898 times 119899 being the size of the image Let each block berepresented as 119861

119896119897(119894 119895) where 119894 and 119895 refer to 119894th row and 119895th

column respectively inside each block Let 119868max119896119897 and 119868min119896119897be the maximum andminimum intensities in an image block119861119896119897(119894 119895)

41 Measure of Enhancement (EME) The EME is defined as

EME119903119888=

1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln(119868max119896119897

119868min119896119897)] (9)

As defined in (3) the CR of a block 119861119896119897is

CR119896119897

=119868max119896119897

119868min119896119897 (10)

Therefore

EME119903119888=

1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (CR119896119897)] (11)

The EME measure is suitable for images with attributes likenoncomplex segments uniform background in segmentssmall targets in segments nonperiodic pattern in segmentsand little to no randomness in segments

42Measure of Enhancement by Entropy (EMEE) TheEMEEis defined as

EMEE120572119903119888

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897

119868min119896119897)

120572

ln(119868max119896119897

119868min119896119897)]

(12)

International Journal of Biomedical Imaging 3

(a) Large uniform background small targetsno randomness and no periodic patterns

(b) Semiperiodic nonuniform back-ground

(c) Large uniform background nonperi-odic pattern

(d) Periodic nonuniform background (e) Random texture nonperiodic nonuni-form background

(f) Random texture nonperiodic nonuni-form background

(g) Nonperiodic nonuniform back-ground

(h) Highly textured nonperiodic nonuniform back-ground

(i) Little texture nonperiodic non-uniform background

Figure 1 Images with different attributes

As defined in (3) the CR of a block 119861119896119897is

CR119896119897

=119868max119896119897

119868min119896119897 (13)

Therefore

EMEE120572119903119888

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (CR119896119897)120572 ln (CR

119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (CR120572)]

(14)

Therefore EMEEmeasure is the entropy of Contrast Ratio foreach block 119861

119896119897 scaled by 120572 averaged over the entire image

The EMEE measure is suitable for images with attributeslike noncomplex segments or small size complex segmentsnonperiodic patterns in segments and randomness because

4 International Journal of Biomedical Imaging

of the inclusion of entropyThemore the value of 120572 the morethe emphasis on entropy The term 120572 helps to handle morerandomness

43 Logarithmic Michelson Contrast Measure (AME) TheAME is defined as

AME119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (119862119872119896119897

)] (15)

The AME measure is suitable for images with attributes likeperiodic patterns in segments and no randomness in textureIt cannot be used for images with uniform background andalso images with segments of large uniform intensity

44 Logarithmic AME by Entropy (AMEE) The AMEE isdefined as

AMEE119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)

120572

sdot ln(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)]

(16)

Michelson contrast is given as

119862119872

=119868max minus 119868min119868max + 119868min

(17)

Therefore

AMEE120572119903119888

= minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (119862119872119896119897

)120572 ln (119862

119872119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (119862119872119896119897

)120572]

(18)

Therefore AMEE measure is the entropy of Michelsoncontrast for each block 119861

119896119897 scaled by 120572 averaged over the

entire imageThe AMEE measure is suitable for images with the

attributes like surplus randomness in texture and periodicpatterns in segments It cannot handle images with largeuniform background

5 Other Measures of Contrast

Here the first two measures are taken from [7] and the nexttwo from [8]

51 Discrete Gray Level Energy (GLE) The gray level energymeasure basically refers to the way of distribution of graylevels It is given as

GLE =

119896

sum

119894=1

119901 (119894)2 (19)

where GLE represents the gray level energy with 256 binsand 119901(119894) refers to the probability distribution functions of

different gray levels that is the histogram count In specialcase GLE reaches its maximum value of 1 when the imagehas a constant value of gray level

GLEmax = max119896

sum

119894=1

119901 (119894)2 = 1 (20)

52 Relative Entropy (RE) Let 119901 and 119902 be two probabilityfunctions of processing images then RE of 119901 with respectto 119902 is defined as

RE =

119896

sum

119894=1

119901 (119894) log2

119901 (119894)

119902 (119894) (21)

Relative entropy is also called the Kullback Leibler distance

53 Second-Order Entropy (SOE) The first-order entropy isused as enhancement metric as it is less sensitive to noiseand has greater response to gray level step But it has thedrawback of ignoring the correlation among pixels Second-order derivatives enhance noise points much more and theyare more sensitive to a line than a step and to a point than toa line

The SOE as defined in [12] uses a cooccurrence matrix toconsider the spatial correlation Let119860 be an image of size 119909 times

119910with 119871 gray levels then element in the cooccurrencematrix119862 of size 119871 times 119871 which contains the information of correlationof the image is given as

119888119894119895

=

119910minus1

sum

119897=0

119909minus1

sum

119896=0

120595 (119897 119896) (22)

where 119888119894119895

is element at 119894th row and 119895th column of matrix 119862

and

120595 (119897 119896) = 1 if (

119860(119897 119896) = 119894 119860 (119897 119896 + 1) = 119895

andor119860 (119897 119896) = 119894 119860 (119897 + 1 119896) = 119895

)

120595 (119897 119896) = 0 otherwise

(23)

The probability of cooccurrence 119901119894119895is computed as

119901119894119895

=119888119894119895

sum119871minus1

119896=0sum119871minus1

119897=0119888119894119895

(24)

and the SOE is computed as

SOE = minussum

119895

sum

119894

119901119894119895log2(119901119894119895) (25)

54 Edge Content (EC) Let 119860(119909 119910) be an image where 119909

is the row coordinate and 119910 is the column coordinate Thegradient vector at any pixel position (119909 119910) is computed as

120575119860 (119909 119910) = [

119866119909

119866119910

] =

[[[[

[

120597

120597119909119860 (119909 119910)

120597

120597119910119860 (119909 119910)

]]]]

]

(26)

International Journal of Biomedical Imaging 5

Table 1 Average metrics values for brain cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 85517 265507 663417 1097868 1845359PSNR mdash 388112 338924 299151 277284 254730AMBE mdash 26961 47250 74828 95790 124371EME 40391 43321 45948 49485 52719 56941EMEE 00219 00236 00251 00272 00292 00317AME 698246 686713 678564 665659 656326 642347AMEE 02166 02138 02118 02087 02063 02029GLE 05199 05226 05265 05289 05331 05363RE mdash 128020 129135 130003 131236 132628SOE 37928 37594 37108 36806 36304 35931EC 09306 09801 10253 10805 11305 11884

where

119866119909= 119860 (119909 119910) minus 119860 (119909 + 1 119910)

119866119910= 119860 (119909 119910) minus 119860 (119909 119910 + 1)

(27)

Taking absolute value of image gradient as an indicator ofcontrast

|120575119860| = radic1198662119909+ 1198662119910 (28)

The measure EC is defined as

EC =1

119903 times 119888sum

119909

sum

119910

1003816100381610038161003816120575119860 (119909 119910)1003816100381610038161003816 (29)

where 119903 times 119888 is the size of image and

1 le 119909 le 119903

1 le 119910 le 119888

(30)

Contrast changes exist over the entire image and EC takesinto account all the contrast changes even for pixels whichare very close (adjacent) to each other

6 Enhancement Procedure

To decide the most appropriate contrast measure anenhancement procedure is required which enhances theimage contrast to different levels For this purpose a Matlabfunction ldquoimadjust()rdquo [13] is used to increase the contrastto different levels First an image is read and if required

converted to gray image Let us call this image the origi-nal image From the original image an adjusted image ofreasonably poor contrast is obtained Let us call this imagereference image The reference image is created using theMatlab command as follows

Reference Image RI = imadjust(Original Image[02 08] [02 08])

where the second parameter that is the input range of graylevels gets mapped to the third parameter that is the outputrange of gray levelsThe full range of gray levels [0 1] for bothinput and output refers to [0 255]

Now using the reference image RI different images ofdifferent levels of enhancement are obtained as follows

1198641 = imadjust(RI [021 079] [019 081])1198642 = imadjust(RI [022 078] [018 082])1198643 = imadjust(RI [023 077] [017 083])1198644 = imadjust(RI [024 076] [016 084])1198645 = imadjust(RI [025 075] [015 085])

where 1198641 1198642 1198643 1198644 and1198645 are five enhanced versions ofthe reference image RIThe enhancements of images are suchthat the contrast of RI lt 1198641 lt 1198642 lt 1198643 lt 1198644 lt 1198645 It canbe seen that for every next level of enhancement the inputrange has been narrowed down and the output range has beenexpanded resulting in contrast stretching and therefore morecontrast

7 Tests and Results

For the purpose of testing twenty medical images are takenfrom [14] ten brain cancer images and ten breast cancerimages From each of these images six versions namely onereference image (RI) and five enhanced images (1198641 to 1198645)are created as discussed above One example of each categorynamely brain cancer and breast cancer is shown in Figures 2and 3 Then the average values of the contrast enhancementmeasures or metrics as mentioned above are computed andthe results are shown in Tables 1 and 2 As far as the complexmeasures EME EMEE AME and AMEE are concerned theimages are divided into blocks of size 16 times 16 and in caseof EMEE and AMEE the value for 120572 is taken as 01 as therandomness in the images of brain and breast cancer is quitelow

Defining contrast stretching (CS) quantitatively withrespect to the reference image (RI) for the different enhancedversions of the images is as given in the following

CS =((HighOut minus LowOut) (HighIn minus LowIn)) of Enhanced Image((HighOut minus LowOut) (HighIn minus LowIn)) of Reference Image

(31)

where [LowOut HighOut] and [LowIn HighIn] are outputrange and input range of gray levels of the image Thedenominator of (31) is 1 since

HighOut minus LowOut

HighIn minus LowInof Reference Image = 08 minus 02

08 minus 02

= 10

(32)

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Biomedical Imaging 3

(a) Large uniform background small targetsno randomness and no periodic patterns

(b) Semiperiodic nonuniform back-ground

(c) Large uniform background nonperi-odic pattern

(d) Periodic nonuniform background (e) Random texture nonperiodic nonuni-form background

(f) Random texture nonperiodic nonuni-form background

(g) Nonperiodic nonuniform back-ground

(h) Highly textured nonperiodic nonuniform back-ground

(i) Little texture nonperiodic non-uniform background

Figure 1 Images with different attributes

As defined in (3) the CR of a block 119861119896119897is

CR119896119897

=119868max119896119897

119868min119896119897 (13)

Therefore

EMEE120572119903119888

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (CR119896119897)120572 ln (CR

119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (CR120572)]

(14)

Therefore EMEEmeasure is the entropy of Contrast Ratio foreach block 119861

119896119897 scaled by 120572 averaged over the entire image

The EMEE measure is suitable for images with attributeslike noncomplex segments or small size complex segmentsnonperiodic patterns in segments and randomness because

4 International Journal of Biomedical Imaging

of the inclusion of entropyThemore the value of 120572 the morethe emphasis on entropy The term 120572 helps to handle morerandomness

43 Logarithmic Michelson Contrast Measure (AME) TheAME is defined as

AME119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (119862119872119896119897

)] (15)

The AME measure is suitable for images with attributes likeperiodic patterns in segments and no randomness in textureIt cannot be used for images with uniform background andalso images with segments of large uniform intensity

44 Logarithmic AME by Entropy (AMEE) The AMEE isdefined as

AMEE119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)

120572

sdot ln(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)]

(16)

Michelson contrast is given as

119862119872

=119868max minus 119868min119868max + 119868min

(17)

Therefore

AMEE120572119903119888

= minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (119862119872119896119897

)120572 ln (119862

119872119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (119862119872119896119897

)120572]

(18)

Therefore AMEE measure is the entropy of Michelsoncontrast for each block 119861

119896119897 scaled by 120572 averaged over the

entire imageThe AMEE measure is suitable for images with the

attributes like surplus randomness in texture and periodicpatterns in segments It cannot handle images with largeuniform background

5 Other Measures of Contrast

Here the first two measures are taken from [7] and the nexttwo from [8]

51 Discrete Gray Level Energy (GLE) The gray level energymeasure basically refers to the way of distribution of graylevels It is given as

GLE =

119896

sum

119894=1

119901 (119894)2 (19)

where GLE represents the gray level energy with 256 binsand 119901(119894) refers to the probability distribution functions of

different gray levels that is the histogram count In specialcase GLE reaches its maximum value of 1 when the imagehas a constant value of gray level

GLEmax = max119896

sum

119894=1

119901 (119894)2 = 1 (20)

52 Relative Entropy (RE) Let 119901 and 119902 be two probabilityfunctions of processing images then RE of 119901 with respectto 119902 is defined as

RE =

119896

sum

119894=1

119901 (119894) log2

119901 (119894)

119902 (119894) (21)

Relative entropy is also called the Kullback Leibler distance

53 Second-Order Entropy (SOE) The first-order entropy isused as enhancement metric as it is less sensitive to noiseand has greater response to gray level step But it has thedrawback of ignoring the correlation among pixels Second-order derivatives enhance noise points much more and theyare more sensitive to a line than a step and to a point than toa line

The SOE as defined in [12] uses a cooccurrence matrix toconsider the spatial correlation Let119860 be an image of size 119909 times

119910with 119871 gray levels then element in the cooccurrencematrix119862 of size 119871 times 119871 which contains the information of correlationof the image is given as

119888119894119895

=

119910minus1

sum

119897=0

119909minus1

sum

119896=0

120595 (119897 119896) (22)

where 119888119894119895

is element at 119894th row and 119895th column of matrix 119862

and

120595 (119897 119896) = 1 if (

119860(119897 119896) = 119894 119860 (119897 119896 + 1) = 119895

andor119860 (119897 119896) = 119894 119860 (119897 + 1 119896) = 119895

)

120595 (119897 119896) = 0 otherwise

(23)

The probability of cooccurrence 119901119894119895is computed as

119901119894119895

=119888119894119895

sum119871minus1

119896=0sum119871minus1

119897=0119888119894119895

(24)

and the SOE is computed as

SOE = minussum

119895

sum

119894

119901119894119895log2(119901119894119895) (25)

54 Edge Content (EC) Let 119860(119909 119910) be an image where 119909

is the row coordinate and 119910 is the column coordinate Thegradient vector at any pixel position (119909 119910) is computed as

120575119860 (119909 119910) = [

119866119909

119866119910

] =

[[[[

[

120597

120597119909119860 (119909 119910)

120597

120597119910119860 (119909 119910)

]]]]

]

(26)

International Journal of Biomedical Imaging 5

Table 1 Average metrics values for brain cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 85517 265507 663417 1097868 1845359PSNR mdash 388112 338924 299151 277284 254730AMBE mdash 26961 47250 74828 95790 124371EME 40391 43321 45948 49485 52719 56941EMEE 00219 00236 00251 00272 00292 00317AME 698246 686713 678564 665659 656326 642347AMEE 02166 02138 02118 02087 02063 02029GLE 05199 05226 05265 05289 05331 05363RE mdash 128020 129135 130003 131236 132628SOE 37928 37594 37108 36806 36304 35931EC 09306 09801 10253 10805 11305 11884

where

119866119909= 119860 (119909 119910) minus 119860 (119909 + 1 119910)

119866119910= 119860 (119909 119910) minus 119860 (119909 119910 + 1)

(27)

Taking absolute value of image gradient as an indicator ofcontrast

|120575119860| = radic1198662119909+ 1198662119910 (28)

The measure EC is defined as

EC =1

119903 times 119888sum

119909

sum

119910

1003816100381610038161003816120575119860 (119909 119910)1003816100381610038161003816 (29)

where 119903 times 119888 is the size of image and

1 le 119909 le 119903

1 le 119910 le 119888

(30)

Contrast changes exist over the entire image and EC takesinto account all the contrast changes even for pixels whichare very close (adjacent) to each other

6 Enhancement Procedure

To decide the most appropriate contrast measure anenhancement procedure is required which enhances theimage contrast to different levels For this purpose a Matlabfunction ldquoimadjust()rdquo [13] is used to increase the contrastto different levels First an image is read and if required

converted to gray image Let us call this image the origi-nal image From the original image an adjusted image ofreasonably poor contrast is obtained Let us call this imagereference image The reference image is created using theMatlab command as follows

Reference Image RI = imadjust(Original Image[02 08] [02 08])

where the second parameter that is the input range of graylevels gets mapped to the third parameter that is the outputrange of gray levelsThe full range of gray levels [0 1] for bothinput and output refers to [0 255]

Now using the reference image RI different images ofdifferent levels of enhancement are obtained as follows

1198641 = imadjust(RI [021 079] [019 081])1198642 = imadjust(RI [022 078] [018 082])1198643 = imadjust(RI [023 077] [017 083])1198644 = imadjust(RI [024 076] [016 084])1198645 = imadjust(RI [025 075] [015 085])

where 1198641 1198642 1198643 1198644 and1198645 are five enhanced versions ofthe reference image RIThe enhancements of images are suchthat the contrast of RI lt 1198641 lt 1198642 lt 1198643 lt 1198644 lt 1198645 It canbe seen that for every next level of enhancement the inputrange has been narrowed down and the output range has beenexpanded resulting in contrast stretching and therefore morecontrast

7 Tests and Results

For the purpose of testing twenty medical images are takenfrom [14] ten brain cancer images and ten breast cancerimages From each of these images six versions namely onereference image (RI) and five enhanced images (1198641 to 1198645)are created as discussed above One example of each categorynamely brain cancer and breast cancer is shown in Figures 2and 3 Then the average values of the contrast enhancementmeasures or metrics as mentioned above are computed andthe results are shown in Tables 1 and 2 As far as the complexmeasures EME EMEE AME and AMEE are concerned theimages are divided into blocks of size 16 times 16 and in caseof EMEE and AMEE the value for 120572 is taken as 01 as therandomness in the images of brain and breast cancer is quitelow

Defining contrast stretching (CS) quantitatively withrespect to the reference image (RI) for the different enhancedversions of the images is as given in the following

CS =((HighOut minus LowOut) (HighIn minus LowIn)) of Enhanced Image((HighOut minus LowOut) (HighIn minus LowIn)) of Reference Image

(31)

where [LowOut HighOut] and [LowIn HighIn] are outputrange and input range of gray levels of the image Thedenominator of (31) is 1 since

HighOut minus LowOut

HighIn minus LowInof Reference Image = 08 minus 02

08 minus 02

= 10

(32)

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

4 International Journal of Biomedical Imaging

of the inclusion of entropyThemore the value of 120572 the morethe emphasis on entropy The term 120572 helps to handle morerandomness

43 Logarithmic Michelson Contrast Measure (AME) TheAME is defined as

AME119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[20 ln (119862119872119896119897

)] (15)

The AME measure is suitable for images with attributes likeperiodic patterns in segments and no randomness in textureIt cannot be used for images with uniform background andalso images with segments of large uniform intensity

44 Logarithmic AME by Entropy (AMEE) The AMEE isdefined as

AMEE119903119888=

minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)

120572

sdot ln(119868max119896119897 minus 119868min119896119897

119868max119896119897 + 119868min119896119897)]

(16)

Michelson contrast is given as

119862119872

=119868max minus 119868min119868max + 119868min

(17)

Therefore

AMEE120572119903119888

= minus1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[120572 (119862119872119896119897

)120572 ln (119862

119872119896119897)]

=1

119903 times 119888

119903

sum

119897=1

119888

sum

119896=1

[Entropy (119862119872119896119897

)120572]

(18)

Therefore AMEE measure is the entropy of Michelsoncontrast for each block 119861

119896119897 scaled by 120572 averaged over the

entire imageThe AMEE measure is suitable for images with the

attributes like surplus randomness in texture and periodicpatterns in segments It cannot handle images with largeuniform background

5 Other Measures of Contrast

Here the first two measures are taken from [7] and the nexttwo from [8]

51 Discrete Gray Level Energy (GLE) The gray level energymeasure basically refers to the way of distribution of graylevels It is given as

GLE =

119896

sum

119894=1

119901 (119894)2 (19)

where GLE represents the gray level energy with 256 binsand 119901(119894) refers to the probability distribution functions of

different gray levels that is the histogram count In specialcase GLE reaches its maximum value of 1 when the imagehas a constant value of gray level

GLEmax = max119896

sum

119894=1

119901 (119894)2 = 1 (20)

52 Relative Entropy (RE) Let 119901 and 119902 be two probabilityfunctions of processing images then RE of 119901 with respectto 119902 is defined as

RE =

119896

sum

119894=1

119901 (119894) log2

119901 (119894)

119902 (119894) (21)

Relative entropy is also called the Kullback Leibler distance

53 Second-Order Entropy (SOE) The first-order entropy isused as enhancement metric as it is less sensitive to noiseand has greater response to gray level step But it has thedrawback of ignoring the correlation among pixels Second-order derivatives enhance noise points much more and theyare more sensitive to a line than a step and to a point than toa line

The SOE as defined in [12] uses a cooccurrence matrix toconsider the spatial correlation Let119860 be an image of size 119909 times

119910with 119871 gray levels then element in the cooccurrencematrix119862 of size 119871 times 119871 which contains the information of correlationof the image is given as

119888119894119895

=

119910minus1

sum

119897=0

119909minus1

sum

119896=0

120595 (119897 119896) (22)

where 119888119894119895

is element at 119894th row and 119895th column of matrix 119862

and

120595 (119897 119896) = 1 if (

119860(119897 119896) = 119894 119860 (119897 119896 + 1) = 119895

andor119860 (119897 119896) = 119894 119860 (119897 + 1 119896) = 119895

)

120595 (119897 119896) = 0 otherwise

(23)

The probability of cooccurrence 119901119894119895is computed as

119901119894119895

=119888119894119895

sum119871minus1

119896=0sum119871minus1

119897=0119888119894119895

(24)

and the SOE is computed as

SOE = minussum

119895

sum

119894

119901119894119895log2(119901119894119895) (25)

54 Edge Content (EC) Let 119860(119909 119910) be an image where 119909

is the row coordinate and 119910 is the column coordinate Thegradient vector at any pixel position (119909 119910) is computed as

120575119860 (119909 119910) = [

119866119909

119866119910

] =

[[[[

[

120597

120597119909119860 (119909 119910)

120597

120597119910119860 (119909 119910)

]]]]

]

(26)

International Journal of Biomedical Imaging 5

Table 1 Average metrics values for brain cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 85517 265507 663417 1097868 1845359PSNR mdash 388112 338924 299151 277284 254730AMBE mdash 26961 47250 74828 95790 124371EME 40391 43321 45948 49485 52719 56941EMEE 00219 00236 00251 00272 00292 00317AME 698246 686713 678564 665659 656326 642347AMEE 02166 02138 02118 02087 02063 02029GLE 05199 05226 05265 05289 05331 05363RE mdash 128020 129135 130003 131236 132628SOE 37928 37594 37108 36806 36304 35931EC 09306 09801 10253 10805 11305 11884

where

119866119909= 119860 (119909 119910) minus 119860 (119909 + 1 119910)

119866119910= 119860 (119909 119910) minus 119860 (119909 119910 + 1)

(27)

Taking absolute value of image gradient as an indicator ofcontrast

|120575119860| = radic1198662119909+ 1198662119910 (28)

The measure EC is defined as

EC =1

119903 times 119888sum

119909

sum

119910

1003816100381610038161003816120575119860 (119909 119910)1003816100381610038161003816 (29)

where 119903 times 119888 is the size of image and

1 le 119909 le 119903

1 le 119910 le 119888

(30)

Contrast changes exist over the entire image and EC takesinto account all the contrast changes even for pixels whichare very close (adjacent) to each other

6 Enhancement Procedure

To decide the most appropriate contrast measure anenhancement procedure is required which enhances theimage contrast to different levels For this purpose a Matlabfunction ldquoimadjust()rdquo [13] is used to increase the contrastto different levels First an image is read and if required

converted to gray image Let us call this image the origi-nal image From the original image an adjusted image ofreasonably poor contrast is obtained Let us call this imagereference image The reference image is created using theMatlab command as follows

Reference Image RI = imadjust(Original Image[02 08] [02 08])

where the second parameter that is the input range of graylevels gets mapped to the third parameter that is the outputrange of gray levelsThe full range of gray levels [0 1] for bothinput and output refers to [0 255]

Now using the reference image RI different images ofdifferent levels of enhancement are obtained as follows

1198641 = imadjust(RI [021 079] [019 081])1198642 = imadjust(RI [022 078] [018 082])1198643 = imadjust(RI [023 077] [017 083])1198644 = imadjust(RI [024 076] [016 084])1198645 = imadjust(RI [025 075] [015 085])

where 1198641 1198642 1198643 1198644 and1198645 are five enhanced versions ofthe reference image RIThe enhancements of images are suchthat the contrast of RI lt 1198641 lt 1198642 lt 1198643 lt 1198644 lt 1198645 It canbe seen that for every next level of enhancement the inputrange has been narrowed down and the output range has beenexpanded resulting in contrast stretching and therefore morecontrast

7 Tests and Results

For the purpose of testing twenty medical images are takenfrom [14] ten brain cancer images and ten breast cancerimages From each of these images six versions namely onereference image (RI) and five enhanced images (1198641 to 1198645)are created as discussed above One example of each categorynamely brain cancer and breast cancer is shown in Figures 2and 3 Then the average values of the contrast enhancementmeasures or metrics as mentioned above are computed andthe results are shown in Tables 1 and 2 As far as the complexmeasures EME EMEE AME and AMEE are concerned theimages are divided into blocks of size 16 times 16 and in caseof EMEE and AMEE the value for 120572 is taken as 01 as therandomness in the images of brain and breast cancer is quitelow

Defining contrast stretching (CS) quantitatively withrespect to the reference image (RI) for the different enhancedversions of the images is as given in the following

CS =((HighOut minus LowOut) (HighIn minus LowIn)) of Enhanced Image((HighOut minus LowOut) (HighIn minus LowIn)) of Reference Image

(31)

where [LowOut HighOut] and [LowIn HighIn] are outputrange and input range of gray levels of the image Thedenominator of (31) is 1 since

HighOut minus LowOut

HighIn minus LowInof Reference Image = 08 minus 02

08 minus 02

= 10

(32)

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Biomedical Imaging 5

Table 1 Average metrics values for brain cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 85517 265507 663417 1097868 1845359PSNR mdash 388112 338924 299151 277284 254730AMBE mdash 26961 47250 74828 95790 124371EME 40391 43321 45948 49485 52719 56941EMEE 00219 00236 00251 00272 00292 00317AME 698246 686713 678564 665659 656326 642347AMEE 02166 02138 02118 02087 02063 02029GLE 05199 05226 05265 05289 05331 05363RE mdash 128020 129135 130003 131236 132628SOE 37928 37594 37108 36806 36304 35931EC 09306 09801 10253 10805 11305 11884

where

119866119909= 119860 (119909 119910) minus 119860 (119909 + 1 119910)

119866119910= 119860 (119909 119910) minus 119860 (119909 119910 + 1)

(27)

Taking absolute value of image gradient as an indicator ofcontrast

|120575119860| = radic1198662119909+ 1198662119910 (28)

The measure EC is defined as

EC =1

119903 times 119888sum

119909

sum

119910

1003816100381610038161003816120575119860 (119909 119910)1003816100381610038161003816 (29)

where 119903 times 119888 is the size of image and

1 le 119909 le 119903

1 le 119910 le 119888

(30)

Contrast changes exist over the entire image and EC takesinto account all the contrast changes even for pixels whichare very close (adjacent) to each other

6 Enhancement Procedure

To decide the most appropriate contrast measure anenhancement procedure is required which enhances theimage contrast to different levels For this purpose a Matlabfunction ldquoimadjust()rdquo [13] is used to increase the contrastto different levels First an image is read and if required

converted to gray image Let us call this image the origi-nal image From the original image an adjusted image ofreasonably poor contrast is obtained Let us call this imagereference image The reference image is created using theMatlab command as follows

Reference Image RI = imadjust(Original Image[02 08] [02 08])

where the second parameter that is the input range of graylevels gets mapped to the third parameter that is the outputrange of gray levelsThe full range of gray levels [0 1] for bothinput and output refers to [0 255]

Now using the reference image RI different images ofdifferent levels of enhancement are obtained as follows

1198641 = imadjust(RI [021 079] [019 081])1198642 = imadjust(RI [022 078] [018 082])1198643 = imadjust(RI [023 077] [017 083])1198644 = imadjust(RI [024 076] [016 084])1198645 = imadjust(RI [025 075] [015 085])

where 1198641 1198642 1198643 1198644 and1198645 are five enhanced versions ofthe reference image RIThe enhancements of images are suchthat the contrast of RI lt 1198641 lt 1198642 lt 1198643 lt 1198644 lt 1198645 It canbe seen that for every next level of enhancement the inputrange has been narrowed down and the output range has beenexpanded resulting in contrast stretching and therefore morecontrast

7 Tests and Results

For the purpose of testing twenty medical images are takenfrom [14] ten brain cancer images and ten breast cancerimages From each of these images six versions namely onereference image (RI) and five enhanced images (1198641 to 1198645)are created as discussed above One example of each categorynamely brain cancer and breast cancer is shown in Figures 2and 3 Then the average values of the contrast enhancementmeasures or metrics as mentioned above are computed andthe results are shown in Tables 1 and 2 As far as the complexmeasures EME EMEE AME and AMEE are concerned theimages are divided into blocks of size 16 times 16 and in caseof EMEE and AMEE the value for 120572 is taken as 01 as therandomness in the images of brain and breast cancer is quitelow

Defining contrast stretching (CS) quantitatively withrespect to the reference image (RI) for the different enhancedversions of the images is as given in the following

CS =((HighOut minus LowOut) (HighIn minus LowIn)) of Enhanced Image((HighOut minus LowOut) (HighIn minus LowIn)) of Reference Image

(31)

where [LowOut HighOut] and [LowIn HighIn] are outputrange and input range of gray levels of the image Thedenominator of (31) is 1 since

HighOut minus LowOut

HighIn minus LowInof Reference Image = 08 minus 02

08 minus 02

= 10

(32)

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

6 International Journal of Biomedical Imaging

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 2 Different versions of one brain cancer image

Therefore (31) reduces to

CS =HighOut minus LowOut

HighIn minus LowInof Enhanced Image (33)

Thus CS with respect to RI for each of the enhanced versionsis computed and given in Table 3

Now to find the most appropriate measure the valuesof CS are taken as standard values and correlation betweenCS and each of the contrast metrics is computed and all arethen compared To calculate correlation average values ofthe metrics as given in Tables 1 and 2 are taken and Pearsoncorrelation [15] is used which is given as

119903 =119899 (sum119909119910) minus (sum119909sum119910)

radic[119899sum1199092 minus (sum119909)2] [119899sum1199102 minus (sum119910)

2]

(34)

where 119909 and 119910 are two variables between which the cor-relation 119903 is to be found and 119899 is the number of dataThe two variables here would be CS and one enhancementmeasure For illustration purpose one calculation for MSE(brain cancer) using Table 4 has been shown

Therefore119903

=5 times 5224878 minus (61418 times 3957668)

radic[5 times 7612834 minus 614182] [5 times 5128593 minus 3957668

2]

(35)

or

119903 = 09830 (36)

The value of 119903 ranges from minus10 to +10 When the valueof 119903 moves away from 00 to minus1 or +1 the strength ofassociation between the two variables increases howeverthey are inversely associated if 119903 is negative Table 5 showsthe correlation values of the different contrast enhancementmetrics used against brain and breast cancer images As perTable 5 themost appropriatemeasure for brain cancer imagesis Edge Content (EC) and for breast cancer images EC standsas the second choice with Absolute Mean Brightness Error(AMBE) being the first choice

8 Conclusion with Discussion

From Table 5 we can see that all measures have quite highcorrelation values The conventional measures MSE PSNRand AMBE are more used for restoration purpose and alsothey do not properly reflect the perceived visual quality ofthe imageTherefore if we neglect the conventionalmeasuresEC is the most appropriate measure for both brain and breastcancer images Regarding complex measures EME EMEEAME and AMEE as they are sensitive to image attributesthey are a good choice for any type of image in generalTo find the best among complex measures we proceed with

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Biomedical Imaging 7

(a) Image-RI (b) Image-1198641 (c) Image-1198642

(d) Image-1198643 (e) Image-1198644 (f) Image-1198645

Figure 3 Different versions of one breast cancer image

Table 2 Average metrics values for breast cancer images

RI 1198641 1198642 1198643 1198644 1198645

MSE mdash 90064 304972 739229 1256293 2071439PSNR mdash 385890 332949 294490 271487 249767AMBE mdash 24062 43634 68167 88055 113444EME 60779 64918 68150 72779 76364 81497EMEE 00325 00349 00368 00395 00417 00448AME 543720 535601 532630 523984 520707 511420AMEE 01791 01766 01753 01727 01714 01687GLE 02531 02649 02825 02934 03113 03240RE mdash 89075 93810 96936 100909 104761SOE 65488 64099 62002 60690 58614 57163EC 16252 17027 17643 18450 19088 19889

Table 3 CS for different enhanced versions

1198641 1198642 1198643 1198644 1198645

CS 10690 11429 12222 13077 14

identifying the attributes of the image Actually there aredifferent attributes of images which can be best handled bydifferent measures one measure cannot tackle all attributesequallywellTherefore it is important to come to a conclusionof correct set of attributes the image in question has Then

Table 4 Illustration for calculation of correlation

119909 (CS) 119910 (MSE) 119909119910 1199092

1199102

1069 85517 914177 114276 7313161143 265507 303448 130622 7049401222 66342 810828 149377 4401221308 109787 143568 171008 12053114 184536 258350 196 3405356142 395767 5224878 7612834 5128593The last row of the table shows the summed up values of each column

only a particularmeasure can be suggested In this paper bothbrain and breast cancer images have the following properties

(i) Large uniform background(ii) Random texture in the foreground but no random-

ness in the background(iii) Nonperiodic pattern in the foreground but absence

of any pattern in the background with the possibilityto be considered periodic because the whole back-ground is dark

(iv) Background containing noncomplex segments withthe foreground containing complex segments

(v) Background containing large area of uniform inten-sity which is absent in the foreground

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

8 International Journal of Biomedical Imaging

Table 5 Correlation values

Metrics Brain cancer images Breast cancer imagesMSE 09830 09846PSNR minus09739 minus09708AMBE 09990 09993EME 09992 09989EMEE 09991 09989AME minus09983 minus09987AMEE minus09986 minus09947GLE 09971 09960RE 09987 09967SOE minus09966 minus09955EC 09997 09991

With these attributes any recommendation would only bea compromise Therefore one way to make things simpleis to examine only the foreground of the image as done inmammograms in [6] Therefore we forget the dark portionand consider the whole brain or breast as background andcells tissues lesions and so forth as targets in the foregroundNow the attributes of the image reduce to nonuniformbackground random texture and nonperiodic pattern Thebest recommendation among the complexmeasures for theseattributes is AMEE AMEE cannot handle areas of uniformintensity well and we do have those types of areas but theyare small so we can go ahead with the recommendation

The other aspect is when we consider the whole imagethat is when we consider the dark portion of the images asbackground and brainbreast as foreground then irrespectiveof the attributes present in the image the best among thecomplex measures can be picked up from Table 5 of thispaper by seeing the correlation values EME stands as the bestchoice whereas EMEE is a strong contender

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper

References

[1] Y Zhou K Panetta and S Agaian ldquoHuman visual system basedmammogram enhancement and analysisrdquo in Proceedings of the2nd International Conference on Image Processing Theory Toolsand Applications (IPTA rsquo10) pp 229ndash234 Paris France July2010

[2] K Panetta Y Zhou S Agaian and H Jia ldquoNonlinear unsharpmasking formammogram enhancementrdquo IEEE Transactions onInformation Technology in Biomedicine vol 15 no 6 pp 918ndash928 2011

[3] S S Agaian B Silver and K A Panetta ldquoTransform coefficienthistogram-based image enhancement algorithms using contrastentropyrdquo IEEE Transactions on Image Processing vol 16 no 3pp 741ndash758 2007

[4] K A Panetta E J Wharton and S S Agaian ldquoHuman visualsystem-based image enhancement and logarithmic contrast

measurerdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 38 no 1 pp 174ndash188 2008

[5] V L Jaya and R Gopikakumari ldquoIEM a new image enhance-ment metric for contrast and sharpness measurementsrdquo Inter-national Journal of Computer Applications vol 79 no 9 2013

[6] K Panetta A Samani and S Agaian ldquoChoosing the optimalspatial domain measure of enhancement for mammogramimagesrdquo International Journal of Biomedical Imaging vol 2014Article ID 937849 8 pages 2014

[7] Z Ye H Mohamadian S-S Pang and S Iyengar ldquoImagecontrast enhancement and quantitative measuringof informa-tion flowrdquo in Proceedings of the 6th International Conference onInformation Security and Privacy (WSEAS rsquo07) Tenerife SpainDecember 2007

[8] A Saleem A Beghdadi and B Boashash ldquoImage fusion-basedcontrast enhancementrdquo EURASIP Journal on Image and VideoProcessing vol 2012 article 10 2012

[9] httpwwwlormedruimagesstoriesrakjpg[10] httpmojodailybruincomwp-contentuploads201302

viking2marsgif[11] S-M Guo C-B Li C-W Chen Y-C Liao and J S-H Tsai

ldquoEnlargement and reduction of imagevideo via discrete cosinetransform pair part 2 reductionrdquo Journal of Electronic Imagingvol 16 no 4 Article ID 043007 pp 1ndash9 2007

[12] N R Pal and S K Pal ldquoEntropic thresholdingrdquo Signal Process-ing vol 16 no 2 pp 97ndash108 1989

[13] R Gonzalez R Woods and S Eddins Digital Image ProcessingUsing MATLAB Gatesmark Publishing 2nd edition 2010

[14] httpspubliccancerimagingarchivenetncialoginjsf[15] httpwwwstatisticshowtocomwhat-is-the-pearson-correla-

tion-coefficient

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of