Vous êtes sur la page 1sur 8

ATBlindTWatermarkingTTechniqueTusingTRedundant

TWaveletTTransformTforTCopyrightTProtection
TDr.P.SatyaNarayana1,D.Lahari2,M.Jyosthna3,M.Mahidharreddy4,N.Dillibabu4
T1ProfessorTofTtheTdeptTofTelectronicsTandTcommunicationTengineering,Aits-Tirupathi
T2345
Btechstudents:deptTofTelectronicsTandTcommunicationTengineering,Aits-Tirupathi

Abstract— InT2016,TMakbolTetTal.TadoptedTanotherTembeddingTt
TATdigitalTwatermarkingTtechniqueTisTanTalternativeTmetho echniqueTderivedTfromTLaiTscheme,TwhereTaTwatermarkT
dTtoTprotectTtheTintellectualTpropertyTofTdigitalTimages.TT isTinsertedTbyTexaminingTtheT1stTcolumnTofTUTobtained
hisTpaperTpresentsTaThybridTblindTwatermarkingTtechnique TfromTtheThybridTDCTT-
TformulatedTbyTcombiningTRDWTTwithTSVDTconsideringT SVDTscheme.TFurthermore,TtheTauthorsTpresentedTblock-
aTtrade- basedTDWT-
offTbetweenTimperceptibilityTandTrobustness.TWatermarkTe
SVDTbasedTonTHVSTcharacteristicsTwhileTtheirTschemeT
mbeddingTlocationsTareTdeterminedTusingTaTmodifiedTentro
providesTdrawbacksTofTtheTdown-
pyTofTtheThostTimage.TWatermarkTembeddingTisTemployed
samplingTofTitsTbands.TTheTshiftTvarianceTofTDWTTprod
TbyTexaminingTtheTorthogonalTmatrixTUTobtainedTfromTth
eThybridTschemeTRDWT-
ucesTinaccuracyTinTwatermarkTextractionTTheyTalsoTprov
SVD.TInTtheTproposedTscheme,TtheTwatermarkTimageTinTb idedTaTfeatureTofTencryptionTforTselectedTblockTlocations
inaryTformatTisTscrambledTbyTArnoldTchaoticTmapTtoTpro .TEncryptedTcoordinatesTonTtheTselectedTblocksTprovided
videTextraTsecurity.TOurTschemeTisTtestedTunderTdifferentT TweakTsecurityTandTconfidentialityTofTtheTwatermarkTim
typesTofTsignalTprocessingTandTgeometricalTattacks.TTheTte age.TItTcanTbeTobtainedTbyTattacksTbyTfindingTtheTsigni
stTresultsTdemonstrateTthatTtheTproposedTschemeTprovides ficantTinformationTofTtheTimage.TFurthermore,TtheyTused
ThigherTrobustnessTandTlessTdistortionTthanTotherTexisting TaTthresholdTforTwatermarkTembeddingTinTtheTorthogona
TschemesTinTwithstandingTJPEG2000Tcompression,Tcropping lTmatrixTUTthatTresultedTinTanTimportantTissueTdueTtoTi
,TscalingTandTotherTnoises. nvalidTimperceptibilityTandTrobustnessTresults.TMoreTcruc
ially,TtheTresultsTinTwereTweakTbecauseTtheTwatermarkT
Keywords— embeddingTusesTtheTsameTthresholdTforTdifferentThybrid
blindTwatermarkingTtechnique;TmodifiedTentropy;TwatermarkTi TschemesT(e.g.TDCT-SVDTandTDWT-
nsertion;TwatermarkTextraction;TredundantTwaveletTtransform SVD).TATthresholdTcannotTbeTusedTforTdifferentThybridT
schemesTandTitTmustTconsiderTaTbalanceTbetweenTimper
I. INTRODUCTION
ceptibilityTandTrobustnessTofTwatermarkedTimages.
WithTtheTadvancementTofTtechnology,TtransmissionTan
dTdistributionTofTdigitalTmultimediaTdataTbecomeTvulnera ThisTpaperTpresentsTaT4×4TRDWT-
bleTtoTunauthorizedTduplication.TTherefore,Tnow-a- SVDTimageTwatermarkingTtechniqueTbasedTonTmodifiedT
daysTwatermarkingTtechniquesTthatTcanTverifyTtheTauthen entropyTconsideringTanToptimalTthresholdTvalue.TArnoldT
chaoticTmapTisTusedTtoTscrambleTtheTwatermarkTimageT
ticityTofTtheTdigitalTdataTareTimportantTtoTprotectTtheTin
forTimprovingTsecurityTandTconfidentialityTofTtheTwaterm
tellectualTpropertiesTorTcopyrightsTinTdigitalTimagesT.TSo
arkedTimage.TTheTscrambledTbinaryTwatermarkTisTembed
meTessentialTfeaturesTofTimageTwatermarkingTareTinvisibi
dedTbyTmodifyingTU3,1TandTU4,1TcoefficientsTofTtheTorth
lity,TrobustnessTandTsecurity.TTheTtechnologyTofTdigitalT
ogonalTmatrixTUTobtainedTfromTRDWT-
watermarkingThasTrecentlyTattractedTtoTimproveTtheTrobu
SVDTusingTspecificTrules.TTheTrobustnessTofTwatermarke
stnessTagainstTseveralTtypesTofTattacks.
dTimagesTisTmeasuredTunderTdifferentTtypesTofTgeometri
ManyTwatermarkingTschemesThaveTbeenTpresentedTusi cTandTsignalTprocessingTattacks.TTheTproposedTschemeTi
ngTfrequencyTdomainTandTsingularTvalueTdecomposition. sTdesignedTtoTreduceTtheTpossibilityTofTtheTfalse-
TMakbol- positiveTproblemTinTtheTextractedTwatermark.TFalse-
KhooTpresentedTanTembeddingTschemeTthatTdirectlyTinse positiveTproblemsTexhibitedTinTMakbol-
rtsTtheTwatermarkTintoTtheTsingularTvaluesTSTobtainedTfr KhooTschemeT,TZhangTandTLiT,TRykaczewskiT,TLiu-
omTredundantTwaveletTtransformT(RDWT)- TanTandTLai-TsaiT.
SVD.TLingTetTal.TverifiedTtheTflawTofTfalse-
positiveTproblemTinTMakbol- II. METHODTANDTMATERIALS
KhooTschemeTinTtheTrecoveredTwatermark.TLingTetTal.
revealedTthatTMakbol- A. ArnoldTScrambling
KhooTschemeTisTdependentTonTtheTwatermark’sTUwTandT ArnoldTscramblingTcanTimproveTtheTsecurityTofTwater
VwTorthogonalTmatrices. markedTimageTArnoldTscramblingTtransformationTisTdefin
edTby:
𝑥 1𝑇1 𝑥
(𝑦)=( ) ( )modTNT(1)TljT[i]T=T(kj+1T[i]*gjT[-i])T(5)T
1𝑇2 𝑦
𝑥 𝑥
where( )TrepresentsTtheTvectorTpositionTafterTshifting( )TRDWTTSynthesis:
𝑦 𝑦
denotesTtheToriginalTvectorTpositionTbeforeTshiftingTandTmodTKj+1[i]T=(Kj[i]*hj[i]+lj[i]*gj[i])T(6)
TstandsTforTtheTmodulusToperationTwithTtheTdivisorTN.TheT
numberTofTscramblingTiterationsTisTalsoTgivenTbyTNT.InTthisTwhereTh[-i]TandTg[-
i]TreferTtoTlowTpassTandThighTpassTanalysisT
paper,NTisTconsideredTasTaTkey.TheTinverseTArnoldTScramblingTfilters,respectively;Th[i]TandTg[i]TdenoteTlow-
passTandThigh-passT
isTdefinedTby:TsysnthesisTfilters;TkjTandTijTrepresentTlowTbandTandThighTband
𝑥 2𝑇 − 1 𝑥
T(𝑦)T=( ) ( )modTNT(2)Tcoefficients,respectivelyTatTlevelTj.TLLTsub-bandTobtainedTfromT
−1𝑇1𝑇 𝑦
TtheTselectedTblockTofTRDWTTcoefficientsTisTthenTdecomposedTbyT
BT.ModifiedTEntropy:TSVD
EntropyTisTmeasuredTbyTtheTspatialTcorrelationTofTthe
NeighbourTpixelsTandTitTcarriesTimportantTinformationTofTanTD.TSingularTValueTDecomposition(SVD):
Image.Meanwhile,edgeTentropyTprovidesTimportantTinformationT
OfTtheTimageTcharacterstics.TheTcombinationTofTtheseTtechniquesTSVDTisTaTwidelyTusedTinTImage-
WaterTmarkingTapplications
IndicatesTlessTdistortionTareaTofTtheTimage.ModifiedTentropyTisTInTSVD,aTmatrixTATisTdecomposedTas
definedTasTfollows:
TA=USVTT(7)TA=USVTT(7)
TWhereTsTisTaTdiagnolTmatrixTwhereTdiagnolTentriesTareTtheT
TN
TSquaresTarrangedTinTlowerTorder;andTUTandTVTareTtheT
ETHVSTT−T∑(TpiTexp1−Tp TTpiTlogT2T(
i (3)TorthonormalTmatrices
TpiT))T/T2 T
Ti1
WhereTpiTdenotesTtheToccuranceTprobabilityTofTithTpixelTwithTIII.T
ProposedTscheme
0T≤piT≤1.TheTblocksTwithTtheTlowestTmodifiedTentropiesTareT
SelectedTforTembeddingTtheTwaterTmarkTA.InsertionTAlgorithims
T
C.TRDWTT:TTheTprocessTofTwatermarkTinsertionTisTdescribedTin
TRDWTTeliminatesTup-samplingTandTdownTsamplingTofTAlgorthimT1TandTfigT1.TATwatermarkTimageTisTembedded
CoefficientsTwhichTappearTinTDWTTduringTfilter-
bankTiterationsTbyTexamingTtheTentriesTofTU3,1TandTU4,1TofTtheT1stTcoloumnTof
AnalysisTandTsynthesisTofTRDWTTcanTbeTpresentedTbyTtheTfollowingTU.TheTrelationshipTisTconsideredTtoTdetermineT
watermark
EquationsTbitTofT0TorT1.
 TSelectedTblock

 StepT3:TDecomposeTLLTsub-bandTofTRDWTT
 TcoefficientsTusingTSVDTintoTU,STandTV.
 T

WatermarkTextraction:
StepT4:TU3,1TandTU4,1TcoefficientsTareTusedTtoT
AlgorithmT1:TInsertionTprocessT TdetermineTtheTwatermarkTbitsTusingTaT
Input:THostTimage,TWatermark;TT=0.055 TruleTasTfollows:
Pre-processing: TIfT,thenTsetTtheTwatermarkT
StepT1:TAThostTimageTisTdividedTintoT8*8Tpixels Tbit=1,else
StepT2:computeTaTmodifiedTentropyTforTeachT TTheTwatermarkTimage.
TNon-overlappingTblock StepT5:TAfterextractionTprocess,applyTtheT
StepT3:selectTblocksTthatThaveTlowestTmodifiedT TinverseTArnoldTtransformTusingTtheTsameT
TEntropyTvaluesTandTsaveTtheTlocationT TkeyTtoTrecoverTtheTwatermakeTimage
TCoordinates.
StepT4:TATbinaryTwatermarkTisTscrambledTusingT Outptut:TWatermarkTextraction
TArnoldTchaoticTusingTaTsecretTkey. T
WatermarkTinsertion:
StepT5:TselectedTblocksTareTtransformedTusingT
TRDWT,theTfirstTlevelTofTLLTsub-bandT
TEXPERIMENTALTRESULTS:
TcoefficientsTisTappliedTbyTSVD.T
StepT6:TTheTfollowingTrulesTareTusedTtoTembedT
ATbinaryTwatermarkTimageTwithT32*32TpixelsTisTshow
TTheTwatermarkTbits:
nTinTFigT3.TArnoldTtransformTisTusedTtoTscrambleTthe
TRuleT1:TifTtheTU3,1TorTU4,1TcoefficientsTareT
TwatermarkTimageTforTimprovingTtheTsecurityTandTconf
TNegativeTvalues,TthenTx=-1,T=-T,
identialityTofTtheTwatermarkTbeforeTinsertion.TourTsche
TOtherwiseTifTtheTU3,1TorTU4,1T
meTwasTtestedTonTsixTbenchmarkTimagesTwithT512*512
TCoefficientsTareTpositive,TWeT TpixelsT,namelyTLena,TSailboat,TPepper,TAirplane,TLake
TSetTx=1,T=T.Then,calculateTthe TandTBaboonTimagesTasTshownTinTFigT4.TestsTinclude
TaverageTU3,1TandTU4,1T TdifferentTtypesTofTsimulatedTattacksTfiltering,Tadjust,Th
TcoefficientsTRuleT2:ifTtheTbinaryTwatermarkTbit=1, istogramTequalization,Tcrop,TrotationT,translationTandTsca
TU3,1=x ling.TComparisonTofTperformanceTofTourTwatermarkingT
TRuleT3:TifTtheTbinaryTwatermarkTbit=0, schemeTisTmadeTwithTanTexistingTscheme.
TU3,1T=x
post-processingTafterTembedding:
stepT7:TApplyTtheTinverseTSVD,thenTperformTtheT
TinverseTRDWTTonTeachTselectedTblock.
Output:WatermarkedTimageTcontainingTaTlogo
B.ExtractionTAlgorithms
TStepsTofTwatermarkTextractionTareTgivenTinTAlgorithm
sT2TandTFigT2.TAlgorithmT2:TExtractionTprocess
Input:TWatermarkedTimage;TselectedTblockTlocations
Pre-processing: 

StepT1:TtheTselectedTblockTlocationsTareTutilizedT 
TtoTextractTtheTWatermark.TSelectedT 
TregionsTareTsplitTintoT8*8Tpixels 
TStepT2:TApplyTtheTfirstTlevelTofTRDWTTonTeachT
whereTrepresentsTexclusiveTORToperation. WT(x,Ty)
TexaminingTtheTcomponents:U3,1TandTU4,1TofTtheT
(a)T(b) orthogonalTmatrixTUTobtainedTfromTRDWT-
Fig.T3.T(a)TATbinaryTwatermarkT32×32Tpixels,T(b)TTheTcorre
spondingTscrambledTwatermark. SVD.watermarkTbitsTareTnotTdirectlyTinsertedTint
oTU.AnToptimalTthresholdTofTMakbolTschemeTan
dTourTschemeTareTshownTinTfigT5TandT6,respect
ively.
T
(a)T(b)T(c)

(d)T(e)T(d)
Fig.T4.T(a)TLenaT(b)TSailboatT(c)TPepperT(d)TAirplaneT(e)
TLakeT(f)TBaboon.

A. ImperceptibilityTMeasurement
TheTWatermarkedTimagesTqualityTisTmeasuredTbyTPSN
RTandTSSIMTmeasurementsTareTusedTtoTevaluateTtheTi
mperceptibilityTandTtheTperceptualTsimilarityTofTaTwater
markedTimage.TheTtermsTPSNRTisTdefinedTas:
TPSNR=10

WhereTgTrefersTtoTtheTwatermarkedTimageTandTfTreprese
ntsTtheThostTimages.TheTsizeTofTtheTwatermarkedTimageT
isTgivenTbyTM*N.TheTtermTSSIMTisTdefinedTby: FigT6.AnToptimalTthresholdTofTMakbolTschemeTbase
SSIM(x,y)=[l(x,y)T.[c(x,y)]T.s[(x,y)] dTonTaTtrade-offTbetweenTSSIMTandTNCTvalues.T
WhereTparametersTwhichTcanTbeTadjusted.ATfurtherTdescri TUsingTtheTexperiment,weTfoundTanToptimalTthreshol
ptionTcanTbeTfoundTinT[23]. dTvalueTasTapproximatelyT0.026TinTtheTembeddingTp
B. RobustnessTMeasurement rocessTofTMakbolTscheme,whileTanToptimalTthreshold
TforTtheTproposedTRDWT-
TRobustnessTpropertiesTofTourTschemeTareTevaluated SVDTschemeTasTaboutT0.055.TheTthresholdTvalueTwa
TbyTnormalisedTcross- sTobtainedTbyTevaluatingTtheTtrade-
correlationTandTBitTerrorTrate.specifically,NCTandTBER offTbetweenTSSIMTandTNCTvaluesTagainstTJPEGTco
mpression.T
TareTusedTtoTcalculateTtheTrobustnessTofTtheTrecove
redTwatermarkTafterTcertainTattackTonTtheTwatermar
kedTimageToccurs.T

WhereTrepresentsTexclusiveTORToperationTW*(x,y)Tdenotes
TtheTwatermarkTextractionTandTW(xTy)TrepresentsTtheTbin
aryTwatermark.
C.TAnToptimalTthresholdTforTinsertionT
TWatermarkTembeddingTisTperformedTbyT
4
.

5
TABLETI.TCOMPARISONTOFTNCTVALUESTFROMTEXTRACTED
WATERMARK

LenaTSailboat
Attack Makbol OurT Makbol OurT
[11] scheme [11] scheme
NoTattackT1.0000T1.0000T1.0000T0.9942
GaussianTLow-passTfilter
0.9192T0.9877T0.9183T0.9781
[3T3]T
GaussianTLow-passTfilter
0.4203T0.9992T0.4378T0.9764
[5T5]T
GaussianTNoiseT0.003T0.8652T0.9347T0.9292T0.9598

GaussianTNoiseT0.03T0.6589T0.7252T0.6816T0.7208

SharpeningT1.0000T1.0000T1.0000T0.9926
Fig.T8.TComparisonTofTNCTcurvesTfromTMakbolTschemeT
MedianTFilterT[3T3]T0.9496T0.9992T0.9434T0.9859 [11]TandTproposedTschemeTagainstTJPEG2000Tcompre
ssion
MedianTFilterT[5T5]T0.0408T0.7103T0.2195T0.7008

PepperTandTSaltT0.2%T0.9925T0.9867T0.9925T0.9858

PepperTandTSaltT2%T0.8673T0.8806T0.9066T0.9038

SpeckleTNoiseT0.03T0.7838T0.8097T0.7580T0.7994

SpeckleTNoiseT0.3T0.5837T0.6301T0.6037T0.6029

PoissonTNoiseT0.9102T0.9522T0.9138T0.9658

AdjustT0.9975T1.0000T0.9983T0.9925
HistogramTEqualization
0.9899T1.0000T0.9992T0.9950
AttackT
JPEGTwithTQFT=40T0.9840T0.9798T0.9823T0.9875

JPEGTwithTQFT=50T0.9747T0.9942T0.9925T0.9892
Fig.T9.TComparisonTofTBERTcurvesTfromTMakbolTschemeT
JPEGTwithTQFT=60T0.9925T1.0000T0.9925T0.9917 [11]TandTproposedTschemeTagainstTJPEG2000Tcompres
TheTproposedTschemeTproducesTlessTdistortionTandTbett sion.
erTperceptualTqualityTthanTtheTexistingTschemesTasTdemo Figs.T8TandT9TshowTtheTcomparisonTbetweenTMakbolT
nstratedTbyTPSNRTandTSSIMTvaluesTinTTableTI.TTheTwa schemeT[11]TandTtheTproposedTschemeTconsideringTcompr
termarkedTimagesTareTtestedTunderTsignalTprocessingTattac essionTwithTJPEG2000TunderTdifferentTcompressionTratios.
ks.TNCTvaluesTofTtheTextractedTwatermarkTafterTapplying TTheTresultsTproveTthatTtheTproposedTschemeTcanTachiev
TdifferentTgeometricTattacksTareTgraphicallyTdepictedTinTF eThigherTrobustnessTthanTMakbolTschemeTwhenTtheTwater
ig.T7.TTheTtestTresultsTdemonstrateTthatTourTschemeTprod markedTimagesTareTconvertedTtoTJPEG2000TwithTaTcomp
ucesThigherTrobustnessTthanTMakbolTschemeT[11]TasTveri ressionTratioTmoreTthanT6.TTheTproposedTschemeTalsoTco
fiedTbyTNCTvaluesTinTTableTII. nfirmsTsuperiorityTinTrobustnessTwithTJPEG2000TwithTdiff
erentTcompressionTratiosTasTverifiedTinTNCTandTBERTval
ues.
TABLETII.TCOMPARISONTOFTVISUALTPERCEPTIONTFROMTEXTRACTED
WATERMARK
LenaTSailboat
Attack TSp Makbol OurTs Makbol
[10] cheme [10]
eckl
NoTattack
eTN
GaussianTLow-
passTFilter oise
[3T3]
GaussianTLow- T0.0
passTFilter 3
[5T5]TGaussianTNoi
seT0.003

SharpeningTMedian

TFilterT[3T3]TPeppe
Fig.T7.TComparisonTofTNCTvaluesTunderTdifferentTtypesTofTg rTandTSaltT0.2%TP
eometricTattack.
epperTandTSaltT2%

6
OurTscheme

7
LenaTSailboat
Attack Makbol OurT Makbol OurT
[10] scheme [10] scheme
PoissonTNoise

Adjust
HistogramTEqualization
Attack
JPEGTwithTQFT=
50

TheTproposedTschemeTproducesTstrongerTresistanceTand
higherTrobustnessTthanTotherTexistingTschemes.TEvenTtheT
watermarkTextractionTisTdamaged,TtheTextractedTwatermar
kTfromTtheTproposedTschemeTcanTstillTbeTobservedTbyTh
umanTvisualTsystems.
V.TCONCLUSION
ThisTresearchTpresentsTaTwatermarkingTschemeTusingTR
DWTTcombinedTwithTSVDTforTprotectingTcopyrights.TOur
TschemeTutilizesTaTmodifiedTentropyTtoTdetermineTtheTe
mbeddingTregionsTwithTlessTdistortion.TATwatermarkTimag
eTisTscrambledTbyTArnoldTtransformTtoTprovideTextraTsec
urityTofTtheTconfidentialTinformation.TTheTscrambledTwate
rmarkTisTembeddedTinTtheThostTimageTbyTexaminingTU3,1
TandTU4,1TcoefficientsTobtainedTfromTRDWT-
SVDTusedTonTtheThostTimage.TOurTschemeTisTtestedTaga
instTdifferentTtypesTofTsignalTprocessingTandTgeometricTat
tacks.TTestTresultsTofTourTschemeTshowTanTimprovement
TinTtermsTofTSSIMTvaluesTandTNCTvaluesTcomparedTtoT
otherTexistingTschemes.TTheTproposedTschemeTconfirmsTs
atisfactoryTresults,Thowever,TourTschemeTrequiresTlittleThi
gherTcomputationalTcostTdueTtoTArnoldTtransformTandTR
DWT.TItTcanTbeTacceptedTbecauseTourTaimTisTtoTimprov
eTtheTrobustnessTagainstTdifferentTtypesTofTattack.
ACKNOWLEDGMENT
TheTauthorsTsincerelyTthankTUniversitiTMalaysiaTPahan
g,TMalaysiaTforTsupportingTthisTresearchTworkTthroughTU
MPTResearchTGrantTSchemeT(RDU170399).
REFERENCES
[1]TN.A.TAbu,TF.TErnawan,TN.TSuryana,TSahibTS,T“ImageTwatermarkin
gTusingTpsychovisualTthresholdToverTtheTedge,”TInformationTandT
CommunicationTTechnology,TICT-EurAsia,Tvol.T7804,Tpp.T519-527,
2013.
[2]TF.TErnawan,T“RobustTimageTwatermarkingTbasedTonTpsychovisualTt
hreshold,”TJournalTofTICTTResearchTandTApplications,Tvol.T10,Tn
o.T3,Tpp.T228-242,T2016.
[3]TF.TErnawan,TM.N.TKabir,TM.TFadliTandTZTMustaffa,T“Block-
basedTTchebichefTimageTwatermarkingTschemeTusingTpsychovisual
Tthreshold,”TInternationalTConferenceTonTScienceTandTTechnology-
ComputerT(ICST
2016),T2016,Tpp.T6-10.
[4]TF.TErnawan,TM.TRamalingam,TA.TS.TSadiq,TZ.TMustaffa,T“AnTimpr
ovedTimperceptibilityTandTrobustnessTofT4x4TDCT-
SVDTimageTwatermarkingTusingTmodifiedTentropy,”TJournalTofTT
elecommunication,TElectronic
andTComputerTEngineering,Tvol.T9,Tno.T2-7,Tpp.T111-
116,T2017.
[

Vous aimerez peut-être aussi