DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
Global Con as Enhancemen Based Image
Fo ensics Using S a is ical Fea u es
Nee u SINGH, Abhina GUPTA, Roop Chand JAIN
Depa men o Elec onics and Communica ion Enginee ing, Jaypee Ins i u e o In o ma ion Technology,
Sec o -62, 201309 Noida, U a P adesh, India
[email p o ec ed], abhina[email p o ec ed], [email p o ec ed]
DOI: 10.15598/aeee. 15i3.2189
Abs ac . The e olu ion o mode n came as, mobile
phones equipped wi h sophis ica ed image edi ing so -
wa e has e olu ionized digi al imaging. In he p o-
cess o image edi ing, con as enhancemen is a e y
common echnique o hide isual aces o ampe ing.
In ou wo k, we ha e employed s a is ical dis ibu ion
o block a iance and AC DCT coe icien s o an im-
age o de ec global con as enhancemen in an image.
The a ia ion in s a is ical pa ame e s o block a iance
and AC DCT coe icien s dis ibu ion o di e en de-
g ees o con as enhancemen a e used as ea u es o
de ec con as enhancemen . An SVM classi ie wi h
10 − old c oss- alida ion is employed. An o e all ac-
cu acy g ea e han 99 % in de ec ion wi h alse a e
less han 2 % has been achie ed. The p oposed me hod
is no el and i can be applied o uncomp essed, p e i-
ously JPEG comp essed and pos enhancemen JPEG
comp essed images wi h high accu acy. The p oposed
me hod does no employ o - epea ed image his og am-
based app oach.
Keywo ds
AC DCT coe icien s, block a iance, con as
enhancemen o ensics.
1. In oduc ion
The ad en o cellula phones wi h high esolu ion and
sophis ica ed came as has ushe ed e olu ion in ou
li es. Addi ionally, digi al de ices a e loaded wi h lo s
o media edi ing and enhancemen so wa es which en-
able common man o play wi h bo h image sou ces as
well as in o ma ion. This has led o he de elopmen
o Digi al Image Fo ensics (DIF), an a ea o esea ch
which a ge s ce i ying he e aci y o images.
One o he key a eas o he DIF is de ec ion o o ge y
in he image is commonly known as ampe ing de ec-
ion. Passi e me hods o DIF in ol e unco e ing digi-
al ampe ing in he absence o any wa e ma k o sig-
na u e inse ed by he came a a he ime o cap u ing
[1], [2], [3], [4] and [5]. Digi al o ge ies al e unde lying
s a is ics o an image. To c ea e a isually impe cep i-
ble modi ica ion, i is equi ed o ma ch ligh ing condi-
ions, e-size, o a e, o s e ch po ions o he images,
e-sa e he inal image ( ypically wi h lossy comp es-
sion such as JPEG), e c. These manipula ions esul in
in oduc ion o speci ic co ela ions in he s a is ics o
images, which on de ec ion can se e as a sign o digi al
ampe ing a de ec o . Fo ma ching o manipula ing
ligh ing condi ions, applica ion o global o local con-
as enhancemen is a e y common and essen ial s ep
o hide isual aces.
In li e a u e, he wo k is mainly ocused on de ec -
ing speci ic ypes o enhancemen , iz., de ec ion o
gamma co ec ion [6] and [7] and median il e ing [8].
In addi ion, hese me hods wo k on he assump ion
ha he ype o enhancemen is known. S amm e .
al. [9] employed his og ams o he de ec ion o global
and local con as enhancemen . In hei app oach,
in oduc ion o he peak and alley due o con as
enhancemen is used as s a is ical signa u e o de ec
o ge y. The majo d awback o his me hod is calcula-
ion o ene gy (pa ame e used o di e en ia e be ween
he enhanced and unenhanced) which equi es manual
selec ion o cu -o equency o sepa a ing low and
high equency egions o il e ou sa u a ed images.
This me hod also ails on JPEG comp essed images.
An imp o ed me hod was p oposed by Xu eng e al.
which elimina es he equi emen o manual selec ion
o he cu -o equency [10]. A me hod o bo h JPEG
comp essed as well as uncomp essed images based on
he gap be ween he bins o his og am is p oposed by
Cao e . al. [11].
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 509
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
In ou wo k, we ha e exploi ed he s a is ical ea u es
o block a iance, and block AC Disc e e Cosine T ans-
o m (DCT) coe icien s o an image, o he pu pose
o con as enhancemen de ec ion. Block a iance is
a a iance o 8×8non-o e lapping blocks o an im-
age. I is known ha he block a iance o an image
ollows he exponen ial amily o dis ibu ions [12]. An
applica ion o a con as enhancemen ope a ion, he
pa ame e s o s a is ical dis ibu ion exhibi a ia ion
ha can be used as a ea u e se o de ec con as
enhancemen . We ha e employed a wo-pa ame e
Gamma dis ibu ion om exponen ial amily [13] o
dis ibu ions o cha ac e ize he block a iance o bo h
unenhanced and con as enhanced images. Fo s a-
is ical modeling o AC DCT coe icien s a composi e
dis ibu ion, Gaussian-Gamma is employed [14]. An
SVM classi ie [15] wi h Gaussian Radial Basis Func-
ion (RBF) is applied o classi y be ween unenhanced
and enhanced images. Fo bo h JPEG comp essed as
well as uncomp essed images, he de ec ion accu acy
o he p oposed me hod is high. The accu acy in de-
ec ion o globally con as enhanced images is mo e
han 99 % wi h alse ala m less han 2 % wi h ew
excep ions.
The o ganiza ion o he pape is as ollows. The ap-
plica ion o s a is ical dis ibu ion pa ame e s as ea-
u es o de ec con as enhancemen is explained in
Sec. 2. The sys em model wi h assump ions and
he de ec ion esul s wi h he me hod o de ec global
con as enhancemen is desc ibed in Sec. 3. The
wo k is concluded in Sec. 4.
2. S a is ical Modeling o
Block Va iance and AC
DCT Coe icien s
The Laplacian dis ibu ion [16], [17] and [18], Gene -
alized Gamma and Gene alized Gaussian dis ibu ion
[19] a e gene ally employed o empi ically model AC
DCT coe icien s o na u al images. Lam e al. [12]
ha e analy ically p o ed ha composi e s a is ical dis-
ibu ions should be employed o model AC DCT co-
e icien s.
2.1. P e ious Wo k: Model o Block
Va iance and AC DCT
Coe icien s o O iginal Image
The s a is ical dis ibu ion o block a iance o a na -
u al image plays an impo an ole in deciding he
composi e dis ibu ion o modeling o AC DCT coe i-
cien s. The dis ibu ions conside ed o block a iance
in li e a u e a e exponen ial, hal -Gaussian, Gamma
and many mo e dis ibu ions [12], [14], [20] and [21].
In ou wo k, we ha e expe imen ally chosen he wo-
pa ame e Gamma dis ibu ion o e o he dis ibu-
ions. The pd o Gamma dis ibu ed block a iance
(σ2) is gi en by:
p(σ2) = (σ2)β−1
αβΓ(β)exp −σ2
α,(1)
whe e α,βand p(σ2)is a scale pa ame e , shape pa-
ame e and pd o block a iance, espec i ely. The
composi e pd o AC DCT coe icien s, p(Iu, )o na -
u al images is gi en by:
p(Iu, ) =
∞
Z
0
p(Iu, /σ2)p(σ2)d(σ2),(2)
whe e Iu, is 8×8block 2D-DCT o an image,
u= 0,1, ...7, = 0,1, ...7, ep esen DCT domain. The
p(Iu, /σ2), is ze o mean Gaussian dis ibu ion [12] de-
ined as:
p(Iu, /σ2) = 1
√2πσ2exp −Iu,
2σ2.(3)
Using Eq. (1), Eq. (2) and Eq. (3), he ob ained
p(Iu, ) o Gamma dis ibu ed block a iance is:
p(Iu, ) = √2(I2
u, /2)(β/2−1/4)
√πΓ(β)α(1/2+β)/2·
·Kβ−1/22qI2
u,
2α,
(4)
whe e K (x)modi ied Bessel unc ion o he hi d kind.
2.2. Modeling Block Va iance and
AC DCT Coe icien s o
Con as Enhanced Images
The applica ion o con as enhancemen o an image
esul s in expansi e o con ac i e mapping o pixel al-
ues in spa ial domain which causes a change in scale
pa ame e o s a is ical dis ibu ion o block a iance.
Addi ionally, change in scale pa ame e o AC DCT
coe icien s is also obse ed. The powe -law ans o -
ma ion which is widely used enhancemen ope a ion,
is de ined as:
ie
x,y = ound 255 ix,y
255γ,(5)
whe e ix,y and ie
x,y,∀x= 0,1...7,y= 0,1...7, ep esen
o iginal and i s enhanced e sion in spa ial domain and
γis powe ac o . By con e ing a non - linea model
in o linea one by aking na u al log, ln, Eq. (5) can be
w i en as:
ln ie
x,y=γln(ix,y) + c. (6)
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 510
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
By employing p ope y o a iance [22], a iance o
8×8block o powe law ans o med images, lσ2
ecan
be de ined as:
lσ2
e(γ∗lix,y +c) = γ2lσ2,(7)
whe e lσ2and lσ2
ea e block a iance o ln o o iginal
and enhanced image. Using Eq. (7) and linea ans-
o ma ion o p(lσ2),p(lσ2
e)becomes:
p(lσ2
e) = (lσ2
e)lβe−1
(lαe)lβeΓ(lβe)exp −lσ2
e
lαe,(8)
whe e lαe,lβeand p(lσ2
e)is a scale pa ame e , shape
pa ame e and pd o block a iance o ln o ie
x,y, e-
spec i ely.
The 8×8block DCT o con as enhanced image is
gi en by:
lIe
u, =γ(lIu, ) + C, (9)
whe e lI ep esen s DCT o ln(i)and C=c·I0,0.
I0,0is a 8×8ma ix which has only one non-ze o alue
and i s i s elemen equals o 8.lIe
u, can be de ined
as a linea combina ion o lIu, and I0,0.
Applying linea ans o ma ion on Eq. (3),
p(lIe
u, /lσ2
e)can be w i en as:
p(lIe
u, /lσ2
e) = 1
p2πlσ2
e
exp −lIe
u, −C2
2lσ2
e!.(10)
By a iable ans o ma ion and subs i u ing Eq. (10)
and Eq. (8) in Eq. (2), pd o DCT coe icien s o con-
as enhanced image o o iginally Gamma dis ibu ed
block a iance becomes:
p(lIe
u, ) = √2((lIe
u, −C)2/2)(lβe/2−1/4)
√πΓ(lβe)(lαe)(1/2+lβe)/2·
·Klβe−1/2
2s(lIe
u, −C)2
2lαe
,
(11)
whe e, Klβe−1/2(x)modi ied Bessel unc ion o he
hi d kind wi h o de lβe−1/2. I is wo h o men ion
ha he scale pa ame e s (lα, lαe) and shape pa ame-
e s (lβ, lβe) o p(lIu, )and p(lIe
u, )a e compu ed us-
ing Nelde -Mead simplex algo i hm o nume ical Max-
imum Likelihood (ML) es ima ion. In ou wo k, we
ha e u ilized h ee AC DCT coe icien s a loca ions
(u= 0, = 1), (u= 1, = 0), and (u= 2, = 0), o
o iginal and enhanced images.
3. Global Con as
Enhancemen Based
Fo ensics
3.1. Sys em Model
Digi al images a e con e ed in o g ayscale o analy-
sis and de ec ion o global con as enhancemen . Fo
p epa a ion o enhanced images, con as enhancemen
is applied in RGB domain only and hen he de ec ion
me hod is applied o a da abase con aining bo h o ig-
inal and enhanced images. The block a iance o ln
o a digi al image is compu ed and i ed o Gamma
dis ibu ion using ML es ima ion app oach. The com-
pu ed scale and shape pa ame e s a e included in he
ea u e se . Also, he image is ans o med om spa-
ial domain in o he equency domain by using he 2D
8×8Block DCT me hod. The dis ibu ion o h ee
AC DCT coe icien s is i ed o composi e Gaussian-
Gamma dis ibu ion. The compu ed scale and shape
pa ame e s o Gaussian-Gamma dis ibu ion a e in-
cluded in he ea u es. Fu he , he mean, a iance,
skewness, ku osis, en opy and ene gy o i ed dis i-
bu ions o block a iance and AC DCT coe icien s a e
compu ed and included in he ea u e se . To cap u e
he addi ional in o ma ion om he da a o block a i-
ance and AC DCT coe icien s due o con as enhance-
men ope a ion, he mean, a iance, skewness and ku -
osis o block a iance and AC DCT coe icien s a e
u he added o he ea u e se . The ea u e se is ap-
plied o Suppo Vec o Machine (SVM) classi ie . Two
da abases UCID [23] and CASIA [24] a e used o his
pu pose. The sys em model o de ec ion o con as
enhancemen is gi en in Fig. 1.
3.2. De ec ion Algo i hm and
Expe imen al Resul s
In he p e ious sec ion, we ha e discussed he o ma-
ion o he ea u e se which esul s in 48 dimensions.
The op imiza ion o he ea u e selec ion is done
expe imen ally. The shape pa ame e o Gamma
and Composi e dis ibu ion was ound independen
o con as enhancemen ope a ion and he e o e,
emo ed om he aining se o SVM, esul ing in
44 dimension ea u es. The esul s o de ec ion a e
exp essed in e ms o T ue Posi i e Ra e (TPR) and
False Posi i e Ra e (FPR). The MATLAB 2016bis
used as a simula ion so wa e. The SVM classi ica ion
wi h ke nel ’RBF’ using 10 − old c oss- alida ion is
employed o con as enhancemen de ec ion. Ou
algo i hm o de ec con as enhancemen is as ollows:
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 511
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
Fig. 1: Sys em model o global con as enhancemen de ec ion.
1. Con e es image om RGB in o g ayscale.
I
x,y =g ayscale I ,(12)
whe e I is a es image.
2. Compu e a iance o block o 8×8o ln o image
and ob ain i s pd .
3. Es ima e pa ame e s o es image, lα by i ing
Gamma dis ibu ion on pd o p(lσ2
).
p(lσ2
) = (lσ2
)lβ −1
(lα )lβ Γ(lβ )exp −lσ2
lα .(13)
4. Compu e h ee AC DCT coe icien s co espond-
ing o (u, ) (0,1),(1,0) and (2,0) om 8×8
block DCT o ln o an image.
5. Es ima e pa ame e s, lα by i ing composi e
Gaussian-Gamma dis ibu ion on pd o p(lI ),lI
is 8×8block DCT o ln o es image.
p(lI ) = √2((lI
u, −C)2/2)(lβI /2−1/4)
√πΓ(lβI )(lαI )(1/2+lβ )/2·
·KlβI −1/2
2s(lI
u, −C)2
2lαI 
.
(14)
6. Calcula e mean, a iance, skewness, ku osis, en-
opy and ene gy o i ed pd o block a iance
and AC DCT coe icien s.
7. Calcula e mean, a iance, skewness and ku osis
o block a iance and AC DCT coe icien s.
8. Combine ea u es a e 44 dimensions.
9. Inpu he ea u es o an SVM ained wi h
10 − old c oss- alida ion o classi y be ween con-
as enhanced o unenhanced images.
The o iginal se o images om UCID and CA-
SIA da abases a e mixed wi h images enhanced wi h
γ= 0.5 o 2and S-Mapping (Sigmoid Func ion) o
o m da abases o 17394 and 10400 uncomp essed im-
ages espec i ely. The ea e , he esul ing da abases
a e JPEG comp essed wi h di e en Quali y Fac o ,
Q = 50, 70, 90, 100. The Recei e Ope a ing Cha -
ac e is ics (ROC) cu es ob ained o o iginal s en-
hanced se (UCID), o iginal s Pos enhancemen
JPEG comp essed UCID (Quali y Fac o , Q = 50,
70, 90, 100) a e shown in Fig. 2(a). The simila
ROC cu es a e also ob ained o CASIA, as shown
in Fig. 2(b).
In Fig. 3, he ROC cu es o UCID and CASIA
da abases o di e en alues o γ= 0.5 o 2a e shown.
Fo UCID da abase (Fig. 3(a)), he o iginal images a e
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 512
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
False posi i e a e
T ue posi i e a e
UCID
Q = 50
Q = 70
Q = 90
Q = 100
TIFF
(a)
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
False posi i e a e
T ue posi i e a e
CASIA
Q = 50
Q = 70
Q = 90
Q = 100
TIFF
(b)
Fig. 2: ROC cu es o con as enhanced images wi h 0.5≤γ≤2and S-Mapping o uncomp essed and JPEG comp essed a
di e en Qs.
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
False posi i e a e
T ue posi i e a e
UCID
γ = 0.5
γ = 0.6
γ = 0.7
γ = 0.8
γ = 0.9
γ = 1.1
γ = 1.2
γ = 1.4
γ = 1.6
γ = 1.8
γ = 2
S−Mapping
(a)
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
False posi i e a e
T ue posi i e a e
CASIA
γ =0.5
γ =0.6
γ =0.7
γ =0.8
γ =0.9
γ =1.1
γ =1.2
γ =1.4
γ =1.6
γ =1.8
γ =2
S−Mapping
(b)
Fig. 3: ROC cu es o di e en alues o γ,0.5≤γ≤2and S-mapping (a) UCID da abase: Uncomp essed (TIFF) images a e
mixed wi h con as enhanced images and images ob ained a e composi e ope a ion o con as enhancemen ollowed
by JPEG comp ession wi h Q = 50, 70, 90 and 100 o each γ alue (b) CASIA da abase: O iginally comp essed (JPEG)
images a e mixed wi h con as enhanced images and images ob ained a e composi e ope a ion o con as enhancemen
ollowed by JPEG comp ession wi h Q = 50, 70, 90 and 100 o each γ alue.
mixed wi h con as enhanced images and images ob-
ained a e composi e ope a ion o con as enhance-
men ollowed by JPEG comp ession wi h Q = 50, 70,
90 and 100 o each γ alue. Fo CASIA da abase
(Fig. 3(b)), he o iginal JPEG images a e also mixed
wi h con as enhanced images and images ob ained
a e composi e ope a ion o con as enhancemen ol-
lowed by JPEG comp ession wi h Q = 50, 70, 90 and
100 o each γ alue.
In Fig. 4, he ROC cu es a e ob ained o he im-
ages (UCID, CASIA, UCID mixed wi h CASIA) con-
as enhanced using powe law ans o ma ion wi h
γ= 0.5 o 2and S-Mapping and he ea e sa ed in
wo o ma s: uncomp essed (TIFF) o ma and JPEG
o ma wi h di e en Qs (Q = 50, 70, 90 and 100).
I may be no ed ha A ea Unde he Cu e (AUC) o
Fig. 2, Fig. 3 and Fig. 4 is g ea e han 98 %.
Ou p oposed me hod achie ed o e all de ec ion e-
sul s as TPR = 99.2 % when FPR = 2 %. Fo UCID
da abase, TPR = 99.3 % when FPR = 2 % and o
CASIA da abase, TPR = 99.2 % when FPR = 2 %.
All hese esul s a e compa able o he me hods o
S amm e . al [9] and Cao e . al [11] o uncomp essed
images as obse ed in ROC cu es o UCID. Howe e ,
S amm’s me hod [9] is limi ed o uncomp essed images
whe eas ou p oposed me hod (Tab. 1) achie ed 99 %
accu acy o o iginally JPEG comp essed images as
shown in ROC cu es o CASIA da abase. The
me hod p oposed by Cao e . al [11] becomes andom
guess when es ed on JPEG comp essed images ha
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 513

DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
Tab. 1: Compa a i e Analysis o Global Con as Enhancemen Fo ensics.
Me hods Accu acy
(TPR, FPR ) Pe o mance Robus ness o Pos
JPEG Comp ession
S amm [9] 98 %,3 % Uncomp essed Images,
High Quali y JPEG Images No
Cao [11] 100 %,1 % Uncomp essed Images,
Low, Medium, and High Quali y JPEG Images Random Guess
P oposed 99 %,2 %
Uncomp essed Images,
JPEG Images,
Pos Enhancemen JPEG Comp essed Images
Robus agains Pos JPEG Comp ession
o all Quali ies
T P R = 98 %,F P R = 3 %
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
False posi i e a e
T ue posi i e a e
UCID
CASIA
UCID+CASIA
Fig. 4: ROC cu es o images (UCID, CASIA, UCID mixed
wi h CASIA) wi h 0.5≤γ≤2, S-mapping and sa ed
in uncomp essed and comp essed o ma s wi h di e en
Qs.
ha e been comp essed a e enhancemen wi h di e en
Qs o bo h o iginally uncomp essed and JPEG com-
p essed images. Ou me hod ou pe o ms he me hod
p oposed by Cao e . al [11] me hod, as shown in Fig. 4.
A de ailed compa ison wi h s a e-o - he-a echniques
is gi en in Tab. 1.
4. Conclusion
In his pape , a no el me hod ha is independen o
image o ma comp essed o uncomp essed is p oposed
o con as enhancemen o ensics. Ou me hod ex-
ploi s a ia ions in s a is ical pa ame e s o he block
a iance and AC DCT Coe icien s o images. Ou
expe imen al esul s ha e shown ha he p oposed
me hod can de ec con as enhancemen in images ob-
ained om composi e ope a ion o con as enhance-
men ollowed by JPEG comp ession wi h high ac-
cu acy, whe eas he s a e-o -a [9] and [11] becomes
a andom guess. Ou me hod has indi idually achie ed
TPR mo e han 97 % wi h FPR less han 3 % in un-
comp essed, 98 % wi h FPR less han 3 % in o iginally
JPEG comp essed and 98 %wi h FPR less han 3 % in
pos enhancemen JPEG comp essed images wi h ew
excep ions. The combine achie ed accu acy is 99 %
when FPR is less han 2 %. This is a adically new
app oach which does no employ image his og am and
is expec ed o open new is as o esea ch a enues in
con as enhancemen based o ensics.
Re e ences
[1] POPESCU, A. C. S a is ical Tools o Digi al Im-
age Fo ensics. Hanno e . 2004. Disse a ion he-
sis. Da mou h College. Supe iso Hany Fa id.
[2] FARID, H. Image o ge y de ec ion.
IEEE Signal P ocessing Magazine. 2009,
ol. 26, iss. 2, pp. 16–25. ISSN 1053-5888.
DOI: 10.1109/MSP.2008.931079.
[3] MAHDIAN, B. and S. SAIC. A bibliog aphy
on blind me hods o iden i ying image o ge y.
Signal P ocessing: Image Communica ion. 2010,
ol. 25, iss. 6, pp. 389–399. ISSN 0923-5965.
DOI: 10.1016/j.image.2010.05.003.
[4] REDI, J. A., W. TAKTAK and J.-L. DUGE-
LAY. Digi al image o ensics: a bookle o be-
ginne s. Mul imedia Tools and Applica ions. 2011,
ol. 51, iss. 1, pp. 133–162. ISSN 1573–7721.
DOI: 10.1007/s11042-010-0620-1.
[5] QAZI, T., K. HAYAT, S. U. KHAN, S. A.
MADANI, I. A. KHAN, J. KOODZIEJ, H. LI, W.
LIN, K. C. YOW and C.-Z. XU. Su ey on blind
image o ge y de ec ion. IET Image P ocessing.
2013, ol. 7, iss. 7, pp. 660–670. ISSN 1751-9659.
DOI: 10.1049/ie -ip .2012.0388.
[6] FARID, H. Blind in e se gamma co ec ion.
IEEE T ansac ion on Image P ocessing. 2001,
ol. 10, iss. 10, pp. 1428–1433. ISSN 1941-0042.
DOI: 10.1109/83.951529.
[7] CAO, G., Y. ZHAO and R. NI. Fo ensic es-
ima ion o gamma co ec ion in digi al im-
ages. In: 17 h IEEE In e na ional Con e ence on
Image P ocessing (ICIP). Hong Kong: IEEE,
2010, pp. 2097–2100. ISBN 978-1-4244-7994-8.
DOI: 10.1109/ICIP.2010.5652701.
[8] YUAN, H.-D. Blind Fo ensics o Median Fil-
e ing in Digi al Images. IEEE T ansac ion
on In o ma ion Fo ensics and Secu i y. 2011,
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 514
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
ol. 6, iss. 4, pp. 1335–1345. ISSN 1556-6021.
DOI: 10.1109/TIFS.2011.2161761.
[9] STAMM, M. C. and K. J. R. LIU. Fo en-
sic de ec ion o image manipula ion using s a-
is ical in insic inge p in s. IEEE T ansac ion
on In o ma ion Fo ensics and Secu i y. 2010,
ol. 5, iss. 3, pp. 492–506. ISSN 1556-6013.
DOI: 10.1109/TIFS.2010.2053202.
[10] LIN, X., X. WEI and C.-T. LI. Two Imp o ed
Fo ensic Me hods o De ec ing Con as En-
hancemen in Digi al Images. In: Media Wa e -
ma king, Secu i y, and Fo ensics. San F ancisco:
SPIE, 2014, pp. 1–10. ISBN 978-0-819-49945-5.
DOI: 10.1117/12.2038644.
[11] CAO, G., Y. ZHAO, R. NI and X. LI.
Con as enhancemen -based o ensics in
digi al images. IEEE T ansac ion on In-
o ma ion Fo ensics and Secu i y. 2014,
ol. 9, iss. 3, pp. 515–525. ISSN 1556-6021.
DOI: 10.1109/TIFS.2014.2300937.
[12] LAM, E. Y. and J. W. GOODMAN. A ma hema -
ical analysis o he DCT coe icien dis ibu ions
o images. IEEE T ansac ion on Image P ocess-
ing. 2000, ol. 9, iss. 10, pp. 1661–1666. ISSN 1941-
0042. DOI: 10.1109/83.869177.
[13] PAL, N., C. JIN and W. K. LIM. Handbook o Ex-
ponen ial and Rela ed Dis ibu ions o Enginee s
and Scien is . 1s ed. New Yo k: CRC P ess, 2006.
ISBN 978-1-584-88138-4.
[14] NADARAJAH, S. Gaussian DCT Coe icien
Models. Ac a Applicandae Ma hema icae. 2009,
ol. 106, iss. 3, pp. 455–472. ISSN 1572-9036.
DOI: 10.1007/s10440-008-9307-2.
[15] DUDA, R. O., P. E. HART and D. G. STORK.
Pa e n Classi ica ion. 2nd ed. New Yo k: Wiley,
2002. ISBN 978-0-471-05669-0.
[16] REININGER, R. and J. GIBSON. Dis ibu ions
o he wo-dimensional DCT coe icien s o im-
ages. IEEE T ansac ion on Communica ion. 1983,
ol. 31, iss. 6, pp. 835–839. ISSN 0090-6778.
DOI: 10.1109/TCOM.1983.1095893.
[17] MULLER, F. Dis ibu ion shape o wo-
dimensional DCT coe icien s o na u al
images. Elec onics Le e s. 1993, ol. 29,
iss. 22, pp. 1935–1936. ISSN 0013-5194.
DOI: 10.1049/el:199931288.
[18] SMOOT, S. R. and L. A. ROWE. DCT co-
e icien dis ibu ions. Human Vision and
Elec onic Imaging. San Jose: SPIE, 1996,
pp. 403–411. ISBN 978-0-8194-2031-2.
DOI: 10.1117/12.238737.
[19] CHANG, J.-H., J. W. SHIN, N. S. KIM
and S. K. MITRA. Image p obabili y dis-
ibu ion based on gene alized gamma unc-
ion. IEEE Signal P ocessing Le e s. 2005,
ol. 12, iss. 4, pp. 325–328. ISSN 1558-2361.
DOI: 10.1109/LSP.2005.843763.
[20] LAM, E. Y. Analysis o he DCT coe icien dis-
ibu ions o documen coding. IEEE Signal P o-
cessing Le e s. 2004, ol. 11, iss. 2, pp. 97–100.
ISSN 1070-9908. DOI: 10.1109/LSP.2003.821789.
[21] NADARAJAH, S. and S. KOTZ. On he DCT
Coe icien Dis ibu ions. IEEE Signal P ocess-
ing Le e s. 2006, ol. 13, iss. 10, pp. 601–603.
ISSN 1558-2361. DOI: 10.1109/LSP.2006.877141.
[22] PAPOULIS, A. and S. U. PILLAI. P obabil-
i y, Random Va iables and S ochas ic P ocesses.
4 h ed. New Delhi: Ta a McG aw-Hill, 2002.
ISBN 978-0-073-66011-0.
[23] SCHAEFER, G. and M. STICH. UCID: an un-
comp essed colo image da abase. S o age and Re-
ie al Me hods and Applica ions o Mul imedia.
San Jose: SPIE, 2004, pp. 1–9. ISBN 978-0-8194-
5275-7. DOI: 10.1117/12.525375.
[24] DONG, J., W. WANG and T. TAN. CASIA Im-
age Tampe ing De ec ion E alua ion Da abase.
In: IEEE China Summi &In e na ional Con-
e ence on In e na ional Con e ence on Signal
and In o ma ion P ocessing (chinaSIP). Beijing:
IEEE, 2013, pp. 422–426. ISBN 978-1-4799-1043-
4. DOI: 10.1109/ChinaSIP.2013.6625374.
Abou Au ho s
Nee u SINGH M.Sc. Nee u Singh ecei ed he
B.Tech. (Elec onics and Communica ion Enginee -
ing) om Inde p as ha Enginee ing College (IPEC),
Technical Uni e si y, Ghaziabad, and M.Tech. (Com-
munica ion Sys ems and Signal P ocessing) om
Jaypee Ins i u e O In o ma ion Technology (JIIT),
Noida. She is pu suing he Ph.D. a JIIT, Noida.
P esen ly wo king as an Assis an P o esso in he
Depa men Elec onics and Communica ion Engi-
nee ing, JIIT, Noida.
Abhina GUPTA D . Abhina Gup a ecei ed
his B.Tech. (Elec ical Eng.) om I.E.T., M.J.P.
Rohilkhand Uni e si y, Ba eilly, India and M.Tech.
(Signal P ocessing) om I.I.T. Guwaha i, India. He
ecei ed his Ph.D. deg ee om School o Compu e
and Sys ems Sciences, Jawaha lal Neh u Uni e si y,
New Delhi in 2013. He was a senio chie enginee a
Samsung Resea ch and De elopmen Ins i u e, Delhi,
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 515
DIGITAL IMAGE PROCESSING AND COMPUTER GRAPHICS VOLUME: 15 |NUMBER: 3 |2017 |SEPTEMBER
India. Cu en ly, he is an Assis an P o esso (S .) in
he Depa men o Elec onics and Communica ion
Enginee ing, JIIT, Noida, India. His esea ch in e es
include signal p ocessing, s a is ical modelling and
machine lea ning. He is an au ho and co-au ho o
many scien i ic publica ions.
Roop Chand JAIN P o . Roop Chand Jain
has been wo king as he Head o Depa men
and a P o esso a he Depa men o Elec onics
and Communica ion Enginee ing, JIIT, Noida, India,
since 2007. He ecei ed his B.Eng. in Elec onics and
Communica ion, M.Eng. in Mic owa es and Rada
om he Uni e si y o Roo kee, Roo kee, M.Eng.
(Elec ical Eng.) and Ph.D. (Elec ical Eng.) om he
Uni e si y o Albe a, Canada in 1988 unde common-
weal h Fellowship P og amme o he Go . o India.
He wo ked in All India Radio (AIR) and Indian Ai -
lines Co po a ion (IA) om 1971–1979 and 1979–1983,
espec i ely. He joined BITS, Pilani, Rajas han, as
a P o esso in Elec ical and Elec onics G oup in
1989.
c
2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 516