Performance Comparison of Digital Image Watermarking Techniques: A Survey

1,249 views

Published on

Digital watermarking is the processing of combined information into a digital signal. A watermark is a secondary image,
which is overlaid on the host image, and provides a means of protecting the image. In order to provide high quality watermarked
image, the watermarked image should be imperceptible. This paper presents different techniques of digital image watermarking based
on spatial & frequency domain, which shows that spatial domain technique provides security & successful recovery of watermark
image and higher PSNR value compared to frequency domain.

Published in: Education
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,249
On SlideShare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
56
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Performance Comparison of Digital Image Watermarking Techniques: A Survey

  1. 1. International Journal of Computer Applications Technology and ResearchVolume 2– Issue 2, 126 - 130, 2013www.ijcat.com 126Performance Comparison of Digital Image WatermarkingTechniques: A SurveyNamita ChandrakarDepartment of Electronics and TelecommunicationShri Shankaracharya Technical CampusBhilai, Indiachandrakarnamita@gmail.comJaspal BaggaDepartment of Information and TechnologyShri Shankaracharya Technical CampusBhilai, Indiabaggajaspal@gmail.comAbstract: Digital watermarking is the processing of combined information into a digital signal. A watermark is a secondary image,which is overlaid on the host image, and provides a means of protecting the image. In order to provide high quality watermarkedimage, the watermarked image should be imperceptible. This paper presents different techniques of digital image watermarking basedon spatial & frequency domain, which shows that spatial domain technique provides security & successful recovery of watermarkimage and higher PSNR value compared to frequency domain.Keywords: Image watermarking, Spatial domain, Frequency domain, Least Significant Bit (LSB), PSNR (Peak Signal to Noise ratio),MSE (Mean Square Error).1. INTRODUCTIONElectronic commerce, commonly known as e-commerce, is the buying and selling of product or service overelectronic systems such as the internet & other computernetworks. So the growth of e-commerce applications in theworld wide web requires the need to increase the security ofdata communications over the internet. To provide security tothese applications data encryption and information hidingtechniques were introduced & developed.There are many approaches like Cryptography,Watermarking and Steganography to transfer the data/imageto the intended user at destination without any modifications[1]. A watermark is a secondary image, which is overlaid onthe primary image, and provides a means of protecting theimage [2].Cryptography was created as a technique forsecuring the secrecy of communication and many differentmethods have been developed to encrypt and decrypt data inorder to keep the message secret. Cryptography only providessecurity by encryption and decryption. There is no protectionafter decryption & only protected content of the messagesbut watermarks can protect content even after they aredecoded.Watermarking is a pattern of bits inserted intodigital image, audio, video or text file that identifies thefile’s copyright information such as author and rights [3].Thus, watermarking is an approach to make sure the data areprotected. Watermarking is designed to be completelyinvisible. Once the watermarking is done, user can send thewatermarked image to other computer so that other user isable to read the watermark or the hidden message in the imageonly if the same algorithm is used. Thus, the watermark canbe protected without being revealed.It may also be necessary to keep the existence of themessage secret. The technique used to implement this, iscalled steganography. Steganography literally means,"covered writing". An ideal steganographic system wouldembed a large amount of information, perfectly securely withno visible degradation to the cover object. Steganographicmethods are in general not robust, that is the hiddeninformation cannot be recovered after data manipulation.Watermarking is robust against attacks. If the existence ofthe hidden information is known it is difficult, ideallyimpossible for an attacker to destroy the embeddedwatermark.Imperceptibility, robustness, inseparability, security,are the features of digital watermarking.2. TYPES OF DIGITAL WATERMARKSWatermarks and watermarking techniques can bedivided into various categories in various ways.1) According to the type of document to be watermarked,watermarking techniques can be divided into fourcategories as follows:i. Text Watermarkingii. Image Watermarkingiii. Audio Watermarkingiv. Video Watermarking2) In other way, the digital watermarks can be divided intothree different types as follows:i. Visible watermark: Visible watermark is a secondarytranslucent overlaid into the primary image.ii. Invisible-Robust watermark: The invisible-robustwatermark is embed in such a way that alternationsmade to the pixel value is perceptually not noticed and itcan be recovered only with appropriate decodingmechanism.iii. Invisible-Fragile watermark: The invisible-fragilewatermark is embedded in such a way that anymodification of the image would alter or destroy thewatermark.A robust watermark should survive a wide varietyof attacks both incidental and malicious [4, 5]. Thesewatermark attacks can be Simple, Detection-disabling,Ambiguity and Removal attacks. Incidental attacks are thosewhich is applied with a purpose other than to destroy thewatermark. Malicious attacks are those which is designed toremove or weaken the watermark.
  2. 2. International Journal of Computer Applications Technology aVolume 2www.ijcat.com3. DIGITAL IMAGE WATERMARKINGFigure 1 shows the gereral block diagram of digitalimage watermarking. Digital Image Watermarking can protectimage, video, audio from unauthorized person, noise,copyright etc. The best known image watermarking methodthat works in the spatial domain is the Least Significant Bit(LSB), which replaces the least significant bits of pixelsselected to hide the information.Figure 1. Block Diagram of Digital Image Watermarking4. PREVIOUS WORKSThe image watermarking techniques can beclassified into two categories:4.1 Spatial-domain techniques (spatialwatermarks) :The spatial-domain techniques directly modify theintensities or color values of some selected pixels.transforms are applied to the host signal during watermarkembedding. Spatial techniques are not very robust againstattacks. The main strengths of pixel domain methods are thatthey are conceptually simple and have very low computationalcomplexities. spatial domain technique, is less timeconsuming as compare to wavelet or frequency domaintechniques.4.1.1 Least Significant Bit (LSB) TechniqueThe simplest spatial-domain image watermarkingtechnique is to embed a watermark in the least significant bits(LSBs) of some randomly selected pixels of the cover image.Example of least significant bit watermarkingImage:10001010 01110100 00011011 01000001 ...Watermark:0 1 1 0 ...Watermarked Image:10001010 01110101 00011011 0100000Schyndel et al. [6] proposed a technique in which awatermark is generated using a m-sequence generator. Thewatermark was embedded to the least significant bit(LSBs)of the original image to produce the watermarked image. Thewatermark was extracted from a suspected image by tthe least significant bits at the proper locations. Detection wasperformed by a cross-correlation of the original and extractedwatermark. They showed that the resulting image containedan invisible watermark with simple extraction procedures. Butthe watermark, was not robust to additive noise.HAJJAJI et al. [7] proposed watermarking of medical image,in which a set of data is inserted in a medical image. Thewatermarking method is based on the least significant bits(LSBs) in order to check the integrity and confidentiality ofmedical information and to maintain confidentiality forpatient and hospital data. For 10% compression rate, theInternational Journal of Computer Applications Technology and ResearchVolume 2– Issue 2, 126 - 130, 2013WATERMARKINGgure 1 shows the gereral block diagram of digitalDigital Image Watermarking can protectimage, video, audio from unauthorized person, noise,copyright etc. The best known image watermarking methode Least Significant Bit(LSB), which replaces the least significant bits of pixelsFigure 1. Block Diagram of Digital Image WatermarkingThe image watermarking techniques can bedomain techniques (spatialdomain techniques directly modify theintensities or color values of some selected pixels. Notransforms are applied to the host signal during watermarks are not very robust againstmethods are thatthey are conceptually simple and have very low computationaldomain technique, is less timeconsuming as compare to wavelet or frequency domain4.1.1 Least Significant Bit (LSB) Techniquedomain image watermarkingtechnique is to embed a watermark in the least significant bits(LSBs) of some randomly selected pixels of the cover image.00011011 01000001 ...1 0 ...01000000 ...sed a technique in which asequence generator. Thethe least significant bit(LSBs)of the original image to produce the watermarked image. Thewatermark was extracted from a suspected image by takingthe least significant bits at the proper locations. Detection wascorrelation of the original and extractedwatermark. They showed that the resulting image containedan invisible watermark with simple extraction procedures. ButMohamed Ali[7] proposed watermarking of medical image,in which a set of data is inserted in a medical image. Thewatermarking method is based on the least significant bitso check the integrity and confidentiality ofmedical information and to maintain confidentiality forpatient and hospital data. For 10% compression rate, thewatermark is successfully recovered. Disadvantage of thesetechnique is that, all the substitutedextract when a Gaussian noise is applied in the watermarkedimage. Puneet Kr Sharma and Rajni, [1] proposed imagewatermarking & different security issues. To hide logo (secretimage) into the cover image they used LSB algorithm. LSBeach of the pixel of the cover image is replaced by the bits ofthe secret image. Then 2ndLSB of each pixel of the coverimage is replaced by the bits of the secret image and so on.Then PSNR and MSE are calculated for different bitsubstitution from LSB to MSB in image. The PSNR and MSEfound for 1stLSB bit substitution was 55.8784respectively. Deepshikha Chopra et al.watermarking technique and a visible watermarking techniqueusing Least Significant Bit (LSB) algorithm,the least significant bits of pixels selected to hide theinformation. They applied various attacks on the watermarkedimage and their impact on quality of images are measuredusing MSE and PSNR. Koushik Palbiomedical image watermarking technique, modified bitreplacement algorithm in spatial domain, which is much betterthan the conventional simple LSB technique. They embeddedmultiple copies of the same information in several bits of thecover image starting from the lower order to the higher orders.So even if some of the information is lost due to an attack ,they still collect the remaining information and recover thewatermark from the cover image using the bit majorityalgorithm. Some author name which works on spatial domainwith their features and results is shown in Table 1.Table 1. SPATIAL DOMAIN TECHNIQUEAUTHORNAMEFEATURESMohamed AliHAJJAJI et al.[7]Data insertion:i) SHA-1(Secure HashAlgorithme)ii) ErrorCorrecting Code(ECC): TurboCodeData detection:Harris CornerDetectorPuneet KrSharma andRajni [1]i) Pseudo-randomnumber generatorii) LSBembeddingalgorithmChopra et al.[8]Least SignificantBit (LSB)algorithm127watermark is successfully recovered. Disadvantage of thesetechnique is that, all the substituted data cannot properlyextract when a Gaussian noise is applied in the watermarkedimage. Puneet Kr Sharma and Rajni, [1] proposed imagewatermarking & different security issues. To hide logo (secretimage) into the cover image they used LSB algorithm. LSB ofeach of the pixel of the cover image is replaced by the bits ofLSB of each pixel of the coverimage is replaced by the bits of the secret image and so on.Then PSNR and MSE are calculated for different bitLSB to MSB in image. The PSNR and MSELSB bit substitution was 55.8784 and 0.1680et al. [8] proposed invisiblewatermarking technique and a visible watermarking techniquegorithm, which replacesthe least significant bits of pixels selected to hide theinformation. They applied various attacks on the watermarkedtheir impact on quality of images are measuredKoushik Pal et al. [9] proposedbiomedical image watermarking technique, modified bitreplacement algorithm in spatial domain, which is much betterthan the conventional simple LSB technique. They embeddedmultiple copies of the same information in several bits of therting from the lower order to the higher orders.So even if some of the information is lost due to an attack ,they still collect the remaining information and recover thewatermark from the cover image using the bit majorityich works on spatial domainwith their features and results is shown in Table 1.SPATIAL DOMAIN TECHNIQUERESULTFor 10%compression rate,the watermark issuccessfullyrecovered (for theIRM andEchographicmedical images).random LSB or 1st BitSubstitutionPSNR = 55.8784& MSE =0.16808th BitSubstitutionPSNR = 14.3467& MSE =2.3900e+003LSB or 1st BitSubstitutionPSNR = 54.87 &MSE = 0.21MSB or 8th BitSubstitutionPSNR = 14.3467&
  3. 3. International Journal of Computer Applications Technology and ResearchVolume 2– Issue 2, 126 - 130, 2013www.ijcat.com 128MSE =2.3900e+003Sharma et al.[16]i)A pseudorandom numbergeneratorii)Theinformationhiding andextraction systemiii)VisualCryptographyiv)Two differentcover images areused for coveringthe secret shareWatermarkedimage forbaboonsPSNR(with oneLSB) (db) =54.45PSNR(with threeLSB) (db)=44.15Watermarkedimage for leenaPSNR(with oneLSB) (db) =51.15PSNR(with threeLSB) (db)=44.174.2 Frequency-domain techniques (spectralwatermarks):The frequency-domain techniques modify thevalues of some transformed coefficients. The frequency-domain technique first transforms an image into a set offrequency domain coefficients. The watermark is thenembedded in the transformed coefficients of the image suchthat the watermark is invisible and more robust for someimage processing operations. Finally, the coefficients areinverse transformed to obtain the watermarked image.Discrete Cosine Transformation (DCT), Discrete FourierTransformation (DFT) and Discrete Wavelet Transformation(DWT) are the three main methods of data transformation.This technique is complex and watermark cannot be easilyrecovered at the receiver end as compared to the spatialdomain technique.Xiang-Gen Xia et al. [10] proposed a watermarkingtechnique based on the Discrete Wavelet Transform (DWT).They performs two–level decomposition using the Haarwavelet filters. The watermark, modeled as Gaussian noise,was added to the middle and high frequency bands of theDWT transformed image. The decoding process involvedtaking the DWT of a potentially marked image. Sections ofthe watermark were extracted and correlated with sections ofthe original watermark. If the cross-correlation was above athreshold, then the watermark was detected. Otherwise, theimage was decomposed into finer and finer bands until theentire. This technique proved to be more robust than the DCTmethod when embedded zero-tree wavelet compression andhalftoning were performed on the watermarked images. MahaSharkas et al. [11] Senior Members IEEE, proposed a dualdigital image watermarking technique for improved protectionand robustness. They applied frequency domain technique(DWT) into the primary watermark image and then embeddedsecondary watermark in the form of a PN sequence. Theresulting image is embedded into the original image to get thewatermarked image. They applied compression, low passfiltering, salt and pepper noise and luminance change attackinto the watermarked image to increase the robustness of thetechnique. In all four attacks secondary watermark wasdetectable. Cheng et al. [12] proposed an algorithm whichwas based on embedding the watermark image in three timesat three different frequency bands, namely, low, medium andhigh and the results proved that the watermark cannot betotally destroyed by either low pass, medium or high passfilter. P.Ramana Reddy et al. [13] proposed an algorithm thatembeds and extracts the watermark in frequency domain andit is checked for salt and pepper & Gaussian noise attacks.They applied watermark in the DWT coefficients of theoriginal image.Iw(x,y)= I(x,y) +k.w(x,y) (1)Where Iw(x,y) represents watermarked image, kdenotes the gain factor. Robustnest of the watermarkedimage increases with the increase in gain k but the quality ofthe final watermarked image is reduced. Preeti Gupta, [14],proposed cryptography based blind image watermarkingtechnique that embed more number of watermark bits in thegray scale cover image. They applied blind watermarkingtechnique that uses watermark nesting and encryption. Anextra watermark is embedded into the main watermark thenmain watermark is embedded into the DWT domain of thecover image. This technique embeds more number of bits inthe cover image. Mistry, [15], proposed digital imagewatermarking and compared different digital watermarkingmethods. Image or video is embedded information data withinan insensible form for human visual system but in a way thatprotects from attacks such as common image processingtechniques. This paper introduced Spatial domain (like LSB)and transform domain (like DCT, DWT) methods. Authorsfound that DCT and DWT watermarking is comparativelymuch better but complex than the spatial domain encoding.Some author name which works on frequency domain withtheir features and results is shown in Table 2.Table 2:FREQUENCY DOMAIN TECHNIQUEAUTHORNAMEFEATURES RESULTHarpuneetKaur [3]i) Watermarknesting (at level 2 ),Means embed onewatermark in otherand encryption.ii) Used DWT basedtechniquePSNR of mainwatermark afterembeddingwatermark1 init = 17.3239 dBPSNR of grayscale coverimage afterembeddingwatermarkedwatermark =37.1587dBXia-mu Niuet al. [17]i)Gray level digitalwatermarkii) Stack filter’sthresholddecompositiontechniqueiii) DCTPSNR = 30.7dBDisadv- due tothe multiplewatermarks, thePSNR of thewatermarkedimage isnotvery highcompared withtraditionalmethod.Xiang-GenXia et al.[10]i)Multiresolutionwatermarkingmethod,They testalgorithm withcommon image
  4. 4. International Journal of Computer Applications Technology and ResearchVolume 2– Issue 2, 126 - 130, 2013www.ijcat.com 129ii)DWT,iii)Pseudo-randomcodes,iv) Haar DWTdistortions.Signature canbe detectedusing DWTcompared toDCT approach.MahaSharkas et al.[11]i)Dual watermarkingtechniqueii) DWT domainPSNR =44.1065dBDisadv-Secondarywatermark wasstill detectablewhen multithreshold DWTtech wasapplied on thewatermarkedimage.Table 1 and Table 2 shows comparision of twodigital image watermarking techniques, which is based onPSNR values and their results. From table it is clear that wecannot embed too much data in the frequency domain becausethe quality of the host image will be distorted significantly.And also complexity of embedding and extracting watermarkin frequency domain is increases and more difficult comparedto spatial domain. Spatial domain provides successfulrecovery of watermark at receiver end.Here different watermarking methods have beenpresented . Most watermarking methods are based on small,pseudorandom changes are applied to selected Coefficients inthe spatial or transform domain. Spatial domain watermarkingschemes are in general less robust toward noise like attacks[5]. But the big advantage of using Spatial domainwatermarking schemes is that the watermark may easily berecovered if the image has been cropped or translated. This isless obvious if the frequency domain is used. Cropping in thespatial domain results in a substantially large distortion in thefrequency domain, which usually destroys the embeddedwatermark. In the case of DCT domain, if DCT blocks arewatermarked, it is important to know the block position forsuccessful watermark decoding. The similar drawbacks in thecase of wavelet domain, because the wavelet transform isneither shift nor rotation invariant. Due to the simplicity andefficiency of the spatial domain, watermark in the spatialdomain is most used.5. CONCLUSIONDifferent techniques of digital imagewatermarking, based on spatial and frequency domaintechniques have been discussed. On the basis of above surveyit is clear that spatial domain is most widely used techniquebecause the watermark can successfully & easily berecovered if the image has been cropped or translated. ascompared to frequency domain.On the other hand frequency domain provides more securitybut at the same time recovery of watermark at the receiverend is more difficult because the complexity increases.Successful recovery of watermark cannot be provided by thefrequency domain techniques.6. RESULTLiterature survey shows that the watermark may beof visible or invisible type and each method has its ownstrengths and weaknesses.The quality of watermarked imagesis measured in terms of PSNR (Peak Signal to Noise Ratio)and MSE (Mean Square Error). In ideal case the value ofPSNR & MSE should be infinite and zero respectively . But itis not possible for watermarked image. So, large PSNR andsmall MSE is desirable.7. REFERENCES[1] Puneet Kr Sharma and Rajni, “Analysis of ImageWatermarking using Least Significant Bit Algorithm”International Journal of Information Sciences and Techniques(IJIST) Vol.2, No.4, July 2012, pp. 95-101.[2] Manpreet Kaur , Sonika Jindal , Sunny Behal “A Study ofDigital Image Watermarking” IJREAS Volume 2, Issue 2(February 2012) ISSN: 2249-3905, pp. 126-136.[3] Harpuneet Kaur, “Robust Image Watermarking Techniqueto Increase Security and Capacity of Watermark Data”Computer Science & Engineering Department, ThaparInstitute of Engineering & Technology. May 2006, pp. 1-69.[4] A. Nikolaidis, S. Tsekeridou, A. Tefas, V Solachidi (Oct.2001), “A survey on watermarking application scenarios andrelated attacks”, IEEE international Conference on ImageProcessing, Vol. 3, pp. 991– 993.[5] Frank Hartung, Martin Kutter, “Multimedia WatermarkingTechniques”, Proceedings ofThe IEEE, July 1999, Vol. 87, No. 7, pp. 1085 – 1103.[6] R. Schyndel, A. Tirkel, and C. Osborne, “A DigitalWatermark,” Proc. IEEE Int. Conf. on Image Processing,Nov. 1994, vol. II, pp. 86-90.[7] Mohamed Ali HAJJAJI Abdellatif MTIBAA El-beyBOURENNANE, “A Watermarking of Medical Image:Method Based "LSB"”, Journal of Emerging Trends inComputing and Information Sciences, VOL. 2, NO. 12,December 2011, ISSN 2079-8407, pp. 714-721.[8] Deepshikha Chopra, Preeti Gupta, Gaur Sanjay B.C., AnilGupta, “Lsb Based Digital Image Watermarking For GrayScale Image” IOSR Journal of Computer Engineering(IOSRJCE) ISSN: 2278-0661, ISBN: 2278-8727 Volume 6,Issue 1 (Sep-Oct. 2012), pp. 36-41.[9] Koushik Pal, Goutam Ghosh, Mahua Bhattacharya, “ AComparative Study between LSB and Modified BitReplacement (MBR) Watermarking Technique in SpatialDomain for Biomedical Image Security” International Journalof Computer Applications and Technology (2278 - 8298)Volume 1 – Issue 1, 2012, pp. 30-39[10] Xiang-Gen Xia, Charles G. Boncelet, and Gonzalo R.Arce, “A Multiresolution Watermark for Digital Images”Proc. IEEE Int. Conf. on Image Processing, Oct. 1997, vol. I,pp. 548-551.[11] Maha Sharkas, Dahlia ElShafie, and Nadder Hamdy,Senior Member IEEE, “A Dual Digital-Image Watermarking
  5. 5. International Journal of Computer Applications Technology and ResearchVolume 2– Issue 2, 126 - 130, 2013www.ijcat.com 130Technique” World Academy of Science, Engineering andTechnology 5 2005, pp. 136-139.[12] Lu, W., Lu, H. and Chung, F.L. (2006) “Robust digitalimage watermarking based on subsampling” AppliedMathematics and Computation, vol. 181, pp. 886-893.[13] P.Ramana Reddy, Munaga. V.N.K. Prasad, D. SreenivasaRao, “Robust Digital Watermarking of Color Images underNoise attacks” International Journal of Recent Trends inEngineering, Vol 1, No. 1, May 2009, pp. 334-338.[14] Preeti Gupta, “Cryptography based digital imagewatermarking algorithm to increase security of watermarkdata” International Journal of Scientific & EngineeringResearch, Volume 3, Issue 9, September 2012 1 ISSN 2229-5518, pp. 1-4.[15] Darshana Mistry, “Comparison of Digital Water Markingmethods,” 21st Computer Science Seminar SA1-T1-7. IJCSE,Vol. 02, No. 09, ISSN : 0975-3397 , 2010, pp. 2905-2909.[16] Mr. Abhay Sharma, Mrs. Rekha Chaturvedi, Mr. NaveenHemrajani, Mr. Dinesh Goyal, “ New Improved and RobustWatermarking Technique based on 3rdLSB substitutionmethod” International Journal of Scientific and ResearchPublications, Volume 2, Issue 3, March 2012 ISSN 2250-3153, pp. 1-4.[17] Xia-mu Niu, Zhe-ming Lu and Sheng-ho Sun, “DigitalWatermarking of Still Images with Gray-Level DigitalWatermarks” Department of Automatic Test and ControlHarbin lnstitute of Technology Harbin, P. R. China, IEEETransactions on Consumer Electronics, Vol. 46, No. 1,February 2000, pp. 137-145.

×