Comparison of Invisible Digital Watermarking Techniques for its Robustness

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072

Comparison of Invisible Digital Watermarking Techniques for its Robustness

1PG Student, VLSI & Embedded Systems, Dept. of ECE, B.N.M Institute of Technology, Bangalore, India 2Professor and Head, Dept. of ECE, B.N.M Institute of Technology, Bangalore, India 3Assistant Professor, Dept. of ECE, B.N.M Institute of Technology, Bangalore, India ***

Abstract An invisible digital watermarking is a process where an image or a video is embeddedwithsomeinformation in the form of a text or an image called as watermark which is not visible to human eye. In this paper, two invisible digital watermarking techniques based on Discrete Cosine Transforms (DCT) and Discrete Wavelet Transforms (DWT) are implemented in MATLAB and are compared for its robustness using correlation coefficient parameter. The measure of this parameter is doneafterperformingalltypesof attacks and degradation on the images. The end results show that invisible image watermarking based on DWT is more robust to all sort of attacks and degradation as compared to DCT.

Key Words: Digital Invisible Watermarking, Discrete Wavelet Transform (DWT), Discrete Cosine Transform (DCT), Robustness, Co efficient Correlation, MATLAB.

1. INTRODUCTION

Duetothewideexpansionofinternetoveryears,availability ofdataondigitalplatformhasincreasedveryrapidly.Oneof themajorproblemistoprotectthemultimediainformation availableandsoastherightfulowneroftheinformationare concernedabouttheirworkbeingillegallyduplicated.Soin ordertomaintainthebalanceofmultimediabeingavailable and also being protected of illegal use, out of many approaches digital watermarking is the one which has gainedalotofinterestintheindustry.

Themainideaofrobustnessofimagesinwatermarkingisto embedthewatermarkwithintheoriginalimagewhichisnot visible for human eye and also protects the images from imageprocessingattacksanddegradations.Inotherwords, themajoraimistodevelopanimagevisibleexactlyassame as the original image, but still allows identification of the hiddenwatermarkwhencomparedwiththekeygivenbythe ownerwhenevernecessary.

This paper is divided into six sections: section 1 gives the introductionofthepaper,section2dealswiththebasicson digitalwatermarking,section3talksaboutimplementation of the whole watermarking process, section 4 talks about implementation of DCT method, section 5 gives an idea aboutimplementationofDWTmethod,section6describes

theresultsoftheworkdoneandsection7istheconclusion oftheworkdone.

2. BASICS ON DIGITAL IMAGE WATERMARKING

The rapid increase in the usage of digital multimedia applicationshascreatedagreatnecessitytogivencopyright protectiontothosedata.Generallywatermarkingisatypeof markerwhichisembeddedinadigitalimagewhichisusedto identifytheownershipoftheimage.Digitalwatermarkingare oftwotypes.1.Spatialdomainmethodwhere,inanimage space,achangeinthepositionofXdirection,willprojectsthe change of position in space.2. Transform domain method, wheretheimageistransformedintofrequencydomainusing differenttransformationsandthenwatermarkingisdone.

A watermark indeed can be described as a unique identificationcodevisibleorinvisiblewhichisembeddedinto the image permanently. Watermarks are of four different types.1.Visiblewatermark:watermarksarevisibleonthe image.2.Invisiblewatermark:watermarksarehiddeninthe imageandnotvisiblebynakedeyes.3.Publicwatermarks: Not so secure watermarks that can be understood and anyone can modify using certain algorithms. 4. Fragile watermark: watermarks that can be destroyed by manipulatingthedata.

3. IMPLEMENTATION

Digital Watermarking has three phases in its life cycle. 1. Embed:hereadigitalimageisembeddedwithawatermark image. 2. Attack: Once the original image is changed, it is more prone to threats and this is known as attack to the system. 3. Protection: it is the detection of the watermark fromanoisewhichmighthavealteredtheoriginalimage.

Consideringthelifecycleofthewatermarkingprocess,Fig 1 showstheflowofthewatermarkingprocessimplementedin this work. The original which has to be marked must be a grayscaleimage,butincaseifacolorimageisselected,itis convertedintoagrayscaleimagebeforeproceedingfurther.

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072

4. DISCRETE COSINE TRANSFORM

TheDCTpermitstobreaktheimageintobandsofdifferent frequency, which makes it much easy to embed the watermarkintothemiddlefrequencybandsoftheimageas showninfigure4.Thesemiddlebandsareselectedasthey don’trepresentthemostimportantvisualpartoftheimages.

Fig -1: Flowchartforwatermarking

The original image used here is an X ray ofa human hand withdimensionsof512x512pixelsasshowninFig 2.The watermarkimageusedhereisabinaryimageofsize64x64 pixelsasshowninFig 3.Whenembeddingandextractingthe watermark,theuserisaskedtoenteracryptographickeyas a password. The two transform domain methods used for digitalinvisiblewatermarkingareDCTandDWT.

Fig 4: 8x8Coefficientstobemodified

4.1 Embedding DCT Watermark

Figure5showstheembeddingflowchartforDCT.Hereone pixeloftheimageishiddeninevery8x8blockofimageby performing2 DDCT.Thenthecoefficientscombinationfor DCTaremodifiedandinverseDCTisperformedtoobtain thewatermarkedimage.

Fig 2: OriginalImage

Fig 3: WatermarkImage

Fig 5: EmbeddingFlowchart

4.2 Extracting DCT watermark

Figure6showstheextractingflowchartforDCT.HereDCT coefficientsofthecorrespondingblocksofanoriginalimage aresubtractedfromtheDCTcoefficientsofthewatermarked imageblock.Ifthesumofdifferencesbetweenthesetwoare

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

greaterthan0,thevalueofthebitdetectedisconsidered1 else0.

diagonalandverticaldetailsandmodifiedatlevels2and3to embedthewatermark.Laterwhichaninverse3levelDWT decomposition is performed to obtain the watermarked image.

Fig 6: ExtractingFlowchart

5. DISCRETE WAVELET TRANSFORM

DWTisawavelettransformmethodwherethewaveletsare sampled discretely. The wavelet transform methods gives the frequency and spatial description of an image. DWT dividestheimageintolowandhighfrequencyparts,where higher frequency parts contain information of edge componentsoftheimage,whilethelowerfrequencyparts areagainsplitintohigherandlowerfrequencybands.The higher frequency bands are most widely used for digital invisiblewatermarkingastheyarelessvisibletohumaneye whenedgesarechanged.

The3leveldecompositionofDWTisasshowninFig 7.

Fig 7: 3leveldecompositionofDWT

5.1 Embedding DWT watermark

Fig 8 shows the embedding flowchart for DWT. After performing 3 level decomposition on both original and watermark images, the coefficients of the horizontal,

Fig -8: EmbeddingFlowchart

5.2 Extracting DWT watermark

Fig 9showstheextractingflowchartforDWT.Here,DWT coefficients of levels 2 and 3 of the original image are subtractedfromeveryDWTcoefficientoflevels2and3of the watermarked image. The differences are added and if their sum’s differences is greater than 0, the value of the detectedbitisconsideredtobe1,else0

Fig 9: ExtractingFlowchart

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal |

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072

6. RESULTS

In this work, both the watermarking methods use the original non marked image for the detection of the watermarkthusmakingthesystemaprivateinvisibledigital watermarking system. Also in this paper, the correlation coefficient which is the similarity measure between the originalwatermarkandextractedwatermarkiscalculated using MATLAB function “corr2”. As this correlation coefficientapproachescloseto1,itisdeterminedthatboth theoriginalandextractedwatermarksarealmostmatching.

Totestthequalityofthedetectedwatermarkwhichiscalled the robustness of the watermarking system, there are a series of attacks performed: JPEG compression, cropping, contrastmodification,brightnessmodification,filteringand addingnoise.Aftereveryattack,a watermark isextracted and is compared with the original watermark using correlationcoefficient.

Whenthewatermarkedimagesarenotpronetoanyattacks, the DCT and DWT techniques yield a good correlation coefficientvalueasshowninTable 1.

Table -1: Correlationcoefficientoforiginalanddetected watermarkwithoutattacks

DCT DWT 0.987946 1

TheJPEGcompressionattacksareperformedfor10qualities andthecorrelationcoefficientvaluesareaslistedinTable 2.ItshowsthatthevaluesincaseoftheDWTmethodsare higher than 0.5 which is up to 20% of JPEG compression quality.HenceDWTmethodismorerobust.

Table 2:Correlationcoefficientoforiginalanddetected watermarkafterJPEGCompressionattack

Quality DCT DWT 10 0.163426 0.318756 20 0.295148 0.587610 30 0.332688 0.610387 40 0.387562 0.725036 50 0.421597 0.836497 60 0.503624 0.803412 70 0.586931 0.901479 80 0.690752 0.966587 90 0.706981 0.975905 100 0.996308 1

Brightness implies the indication of more gray level in an image.Thisattackisdoneover10differentlevelsandthe correlationcoefficientvalues areaslistedinTable 3.Itis observedthatonamajorityDCTmethodismorerobustthan DWT.

Table -3: Correlationcoefficientoforiginalanddetected watermarkafterbrightnessattack

Level DCT DWT

0.1 0.962381 1 0.1 0.967857 0.994860 0.2 0.93785 0.971269 0.2 0.882490 0.921475 0.3 0.819575 0.849625 0.3 0.726942 0.612589 0.4 0.652369 0.603470 0.4 0.606777 0.325871 0.5 0.456297 0.409787 0.5 0.425778 0.214085

Contrastistheratioofminimumandmaximumvaluesofthe graylevelintheimages.Whilehistogramisthedistribution of the gray level over the image. Correlation coefficient valuesfororiginalanddetectedwatermarkwithhistogram equalization contrast attack is listed in Table 4. It isseen thatbothmethodsshowapproximatelysamevaluesforthis attack.

Table 4:Correlationcoefficientoforiginalanddetected watermarkaftercontrastattack

Level DCT DWT

0.5 0.152364 0.251268 0.6 0.180369 0.278569 0.7 0.236187 0.284156 0.9 0.279641 0.281456 1.1 0.286147 0.278945 1.4 0.291025 0.271025 1.5 0.296947 0.276301 1.6 0.300156 0.269785 1.7 0.314789 0.268941 1.8 0.297510 0.087456

International Research Journal of Engineering and Technology (IRJET)

e ISSN: 2395 0056

Croppingisalteringtheimagebyremovingsomepartofit, whiletheremovedpartarepaddedwith0s.Differentlevels ofcroppingisdoneandtheircorrelationcoefficientvalues are listed below in Table 5. It is seen that both methods perform well even after half of the pixels are replaced by zeros.

Table -5: Correlationcoefficientoforiginalanddetected watermarkaftercroppingattack

Level DCT DWT

1.1 0.164259 0.065148

1.2 0.275630 0.154783

1.4 0.469512 0.289641

1.6 0.584102 0.423271

1.8 0.650326 0.539245

2 0.720159 0.674715

4 0.926314 0.970364

8 0.985201 0.983014

16 0.980236 0.995214

Imagefilteringisperformedbyconvolutionofimageandthe impulseresponseofthefilter.Averaging(LFfiltering)and Sharpening(HFfiltering)arealsoperformedontheimage andtheirvaluesarelistedinTable 6.Itisconcludedthaton amajoritybasisDWTismorerobustthanDCTasitdoesn’t tolerateaveraging.

Table 6: Correlationcoefficientoforiginalanddetected watermarkafterfilteringattack

Type DCT DWT

Average 0.186485 0.465120

Gaussian 0.762014 0.920365

Laplace 0.624596 0.892145

Log 0.542360 0.810265

Median 0.317526 0.69785

Unsharp 0.564523 0.804526

Noiseisanunwantedpixelthatdoesn’tbelongtotheoriginal picture.Gaussiannoise,impulsenoise(Salt&pepper)which is characterized by pixels that deviate from their surroundings,and Speckle noise, whichis a multiplicative noise are the different types of noise attacks that are performedonthewatermark.TheirvaluesarelistedinTable 7. It is seen that DWT is more robust except for salt & peppernoiseattacks,wherethesituationisunresolved.

Table 7: Correlationcoefficientoforiginalanddetected watermarkafternoiseattack

Type DCT DWT

Gaussian 0.275630 0.569085

Salt & Pepper 0.542697 0.510697 Speckle 0.302689 0.536478

7. CONCLUSION

Inthispaper,twodifferenttypesofinvisibledigitalimage watermarkingmethodsusingDCTandDWTtechniquesare implemented.TheirRobustnessiscomparedafterdifferent typesofattacksanddegradation.Thecorrelationcoefficient whichisthesimilaritymeasureoftheoriginalandextracted watermarkaretakenasameasureforRobustness.

Considering the obtained results, it is concluded that invisible digital image watermarking based on DWT technique is more robust than the image watermarking basedonDCTtechnique.

REFERENCES

[1] Cox, IJ, Miller, ML & Bloom, JA 2002, "Digital Watermarking", Morgan Kaufmann Publisher, San Francisco,CA,USA2002

[2] Christine I. Podilchuk, Edward J. Delp, ―"Digital watermarking:Algorithmsandapplication",IEEESignal processingMagazine,July2001.

[3] Sumedh P. Ingale, Dr.C.A.Dhote,"A Survey Of Digital WatermarkingTechniques"

[4] Vinita Gupta,Mr.Atul Barve ,"A Review on Image WatermarkingandItsTechniques"

[5] ManpreetKaur,SonikaJindal,SunnyBehal,"ASTUDYOF DIGITALIMAGEWATERMARKING"

[6] Lalit Kumar Saini, Vishal Shrivastava " A Survey of DigitalWatermarkingTechniquesanditsApplications"

[7] G.RoslineNesaKumari,B.VijayaKumar,L.Sumalatha, and Dr V. V. Krishna," Secure and Robust Digital Watermarking on Grey Level Images", International JournalofAdvancedScienceandTechnology,2009

[8] Baisa L. Gunjal, R.R. Manthalkar, "An overview of transform domain robust digital image watermarking algorithms",JournalofEmergingTrendsinComputing andInformationSciences,2010

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal |

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

[9] W. Hong and M. Hang, "Robust Digital Watermarking SchemeforCopyRightProtection,"IEEETrans.Signal Process,2006.

[10] D.S. Bormane, Pallavi Patil, “DWT Based Invisible Watermarking Technique for Digital Images”, International Journal of Engineering and Advanced Technology (IJEAT) , April 2013, ISSN: 2249 8958, Volume 2,Issue 4

[11] DeepshikhaChopra,PreetiGupta,GaurSanjayB.C.,Anil Gupta,“LsbBasedDigitalImageWatermarkingForGray Scale Image” IOSR Journal of Computer Engineering (IOSRJCE)

[12] Qing Liu,Jun Ying(2012),”Grayscale Image Digital WatermarkingTechnologyBasedonWaveletAnalysis”

[13] Mohamed Sathik, SS Sujatha, “A novel DWT based InvisibleWatermarkingTechniqueforDigitalImages”, InternationalArabJournalofe Technology,Jan2,Vol.2, No.3

[14] Li Sukang, Mei Jiansheng, and Tan Xiaomei, “Digital Watermarking Algorithm Based On DCT and DWT”, Proceedings of the 2009 International Symposium on WebInformationSystemsandApplications(WISA’09), China,May22 24,2009,pp.104 107

[15] Ikpyo Hong, Intaek Kim, Seung So0 Han, “A Blind Watermarking Technique Using Wavelet Transform”, ISIE2001,Pusan,KOREA,07803 70~9 2/01,IEEE

[16] PravinM.Pithiya,H.L.Desai,“DCTBasedDigitalImage Watermarking, Dewatermarking & Authentication”, International Journal of Latest Trends in Engineering andTechnology(IJLTET),Issue3May2013,Vol.2

[17] Shahin Shaikh, Manjusha Deshmukh, “Digital Image WatermarkingInDCTDomain,InternationalJournalof Emerging Technology and Advanced Engineering, Volume3,Issue4,April2013

[18] Sharma, Janki Ballabh et al. “Comparison of Bit Error Rate Performance Measure of Robust Watermarking Scheme Based On Self Fractional Fourier Functions , Empirical Mode Decomposition and Cyclic Encoding.” (2013).

Volume: 09 Issue: 07 | July 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal |