隨著網路的風行和燒錄設備數量的增加，這使得數位資料傳遞變得越來越容易、快速。為了去防止一些盜版業者在未經受權的情況下，肆意地複製數位創作，發展了很多防盜版的方法要來解決上面的問題。但是，其中最廣為大眾所使用的防盜技術─密碼學，卻只能保障傳輸時的安全性。一旦這被加密的資料被解密後，資料就不再受到保護。因此，對於這類被解密的數位媒體，有很急迫的需要去發展出另一套技術來防止惡意攻擊者任意散佈被解密的媒體。數位浮水印（Digital Watermarking）正是一個能夠解決上面密碼學缺點的新興技術。無論數位媒體是否加密，只要藉著抽取出這隱藏之訊息，即可輕易的判斷這數位資訊的合法性與否。

數位浮水印有一項獨一無二的特性，即是數位浮水印和數位多媒體是密不可分的。要滿足這項特性意味著浮水印必須要有能力去存活在數位多媒體所有可能處在的環境。而在有限的資源下，壓縮是影音媒體一道必要的處理過程，換言之，壓縮處理是影音媒體的一項重要資料處理的步驟，因此，本論文將從壓縮的觀點來進行數位浮水印的研究。然在眾多壓縮方法之中, JPEG和MPEG2是目前公認的壓縮標準。它們都是基於離散餘弦轉換(DCT)的編碼技術。因此在本論文中，我們的浮水印系統將針對上面這兩種壓縮標準進行設計。

在本篇論文中，我們提出一個可依照影像視訊壓縮及品質調變的強韌浮水印系統。首先，我們先將摻入的領域（Embedded domain）切割成數個區段，並且讓每一個區段去代表一個浮水印的值。整個摻入浮水印過程，就是讓原本我們想摻入的係數的值落在可以正確代表我們浮水印值的區段。在上面這種方法裡，浮水印抵抗壓縮的能力和媒體的觀看品質與每一個區段的長度有非常密切的關係。 於是在我們的系統裡，我們先讓使用者去選擇他們所希望浮水印能夠存活的壓縮強度。在能達到使用者所設定得壓縮強度前提下，我們要儘量的提升整個媒體的觀看品質。為了要達成上面所描述的目的，我們先建立一個能夠掌控每一個值在壓縮後位移量的表，藉著使用這個所求出來的表，我們提供一種搜尋的方法來決定每一個區段的長度。除了決定區段長度之外，對於一般語意學(Semantic)的浮水印，我們也發展出一種方法，這方法可以藉著減少所需要摻入浮水印的數量，來達到增強媒體品質的目的。

在我們的系統中，我們提供兩種模式來讓大眾做選擇。在第一種「簡單」模式裡面，使用者只需要去選擇想要摻入的浮水印並按下摻入的按鈕，就可以順利的把他們的浮水印摻入進去。基本上，這一種摻入模式就已經可以去滿足絕大多數的需求，但如果使用者想要更強大的效果的時候，他們可以使用第二種「進階」模式。在進階模式中，使用者可以依照他們所需要的安全程度，來決定用「對稱」或「非對稱」金鑰管理(key management)來傳遞所使用的參數。

我們用一連串的實驗來測試我們的系統。經由JPEG和MPEG2壓縮，我們發現測試的結果能夠符合我們理論預期。除了壓縮測試之外，我們在不調整任何參數的條件下，對其它數種攻擊做測試，所得出的實驗結果也都能顯示我們的系統是合宜且有效的。

With the popular of internet and the increasing of recorder equipment, the digital media can distribute more easy and faster. In order to avoid the pirator from copying digital multimedia without the real owner's grant, several technologies are developed to deal with this problem. However, traditional cryptography, the most widespread protected method, only can guarantee the security of transmission. Once decrypted, the data is no longer protected. For decrypted media, there is a strong demand to prevent the malicious attacker from distributing the decrypted media arbitrarily. Digital watermarking is a novel technique that can overcome above disadvantage. Whether the media is decrypted or not, we can easily judge the legitimacy by checking the extracted information.

The unique property of the watermark is that watermark should not be separated from the embedded media. That means that the embedded watermark should survive in the circumstance where media may locate. In limited resource, compression is an inevitable processing for media. That is why the watermarking system should be designed from the viewpoint of compression. In numerous compression approaches, JPEG and MPEG2, both DCT-based coding, are the most popular compression standards in this world. In this thesis, we will develop a watermarking scheme which aims for above two compression standards.

We scheme to propose a robust watermarking system which can be adjusted according to image and video compression and quality. At first, the embedded domain such as DCT coefficients is partitioned into several ranges with each range representing a watermark bit. Embedding procedure is to let original coefficient we want to embed falls in the certain point of the region that can correctly represent watermark. The length of every range has very huge impact on the ability in resisting compression and media quality. In our system, the user can choose the compression intensity they want their watermark can survive. For the situation without harming the robustness of watermark under the chosen requirement of compression, we then optimize the quality of the media. In order to achieve this goal, a table which can completely describe the shifted quantity of every value after compression is created. By using this table, a searching algorithm is proposed to determine the length of every range. In addition, for general semantic watermarks, we also develop a method which improves the media quality by reducing the number of watermark bits we practically need to embed.

In the proposed system, two modes, the basic mode and advanced mode, are provided. In the basic mode, the users need only to choose their watermark and press the embedding button. This basic mode can satisfy a great part of applications. If someone needs improving performance, the advanced mode can be used by utilizing the asymmetrical and symmetrical key management in different secure requirement.

A series of experiments were conducted and the results show that our system can fit theoretical anticipating under JPEG and MPEG2 compression. Under the same parameters, we also test the survival ratio of watermark under several attacks except for compression. The experimental results also demonstrate that our system is feasible and effective.

Bibliography
[1]J.A.Bloom,I.J.Cox,T.Kalker,J.P.M.G.Linnartz,ML.Miller,andC. B.S.Traw,“Copyprotectionfordvdvideo,”inProceedings,IEEE,1999,vol.87,pp.1267–1276.
[2] I.J.Cox,M.L.Miller,andJ.A.Bloom,Digitalwatermarking,2001.
[3]C.Y.LinandS.F.Chang,“Arobustimageauthenticationmethoddistin-guishingjpegcompressionfrommaliciousmanipulation,”IEEETransactionsonCircuitsandSystemsforVideoTechnology,vol.11,pp.153–168,2001.
[4]C.K.HoandC.T.Li,“Semi-fragilewatermarkingschemeforauthenti-cationofjpegimages,”inProceedings,2004InternationalConferenceonInformationTechnology:CodingandComputing(ITCC’2004),2004,vol.1,pp.7–11.
[5]Y.Liu,W.Gao,H.Yao,andS.Liu,“Atexture-basedtamperdetectionschemebyfragilewatermark,”inProceedings,2004InternationalSymposiumonCircuitsandSystems,2004,vol.2,pp.177–180.
[6]F.HartungandB.Girod,“Watermarkingofuncompressedandcompressedvideo,”SignalProcessing,vol.66,pp.283–301,1998.
[7]J.J.Eggers,R.Bauml,R.Tzschoppe,andB.Girod,“Scalarcostaschemeforinformationembedding,”IEEETransactionsonSignalProcessing,vol.51,no.4,pp.1003–1019,2003.
[8]P.G.Fernando,F.Balado,andJ.R.H.Martin,“Performanceanalysisofexistingandnewmethodsfordatahidingwithknown-hostinformationinadditivechannels,”IEEETransactionsonSignalProcessing,vol.51,pp.960–980,2003.
[9]P.W.WongandN.Memon,“Secretandpublickeyimagewatermarkingschemesforimageauthenticationandownershipverification,”IEEETrans-actionsonImageProcessing,vol.10,pp.1593–1601,2001.
[10]M.S.Hwang,C.C.Chang,andK.F.Hwang,“Awatermarkingtechniquebasedonone-wayhashfunctions,”IEEETransactionsonConsumerElec-tronics,vol.45,pp.286–294,1999.
[11]R.G.V.Schyndel,A.Z.Tirkel,andC.F.Osborne,“Adigitalwatermark.,”inProceedings,1994IEEE1stInternationalConferenceonImageProcess-ing,1994,pp.86–90.
[12]I.PitasandT.H.Kasklis,“Applyingsignaturesondigitalimages,”inProceedings,1995IEEENonlinearSignalandImageProcessing,1995,pp.460–463.
[13]O.Bruyndonckx,J.J.Quisquater,andB.Macq,“Spatialmethodforcopy-rightlabelingofdigitalimages,”inProceedings,1995IEEENonlinearSignalandImageProcessing,1995,pp.456–459.
[14]G.Caronni,“Assuringownershiprightsfordigitalimages,”inProceedings,ReliableITSystems,1995,pp.251–263.
[15]S.C.HsiaandI.C.Jou,“Ahighrobustwatermarkingtechniqueusingsub-bandfiltering,”inProceedings,200410thInternationalMultimediaMod-ellingConference,2004,pp.72–78.
[16]C.F.WuandW.S.Hsieh,“Digitalwatermarkingusingzerotreeofdct,”IEEETransactionsonConsumerElectronics,vol.46,pp.87–94,2000.
[17]J.Huang,Y.Q.Shi,andY.Shi,“Embeddingimagewatermarksindccom-ponents,”IEEETransactionsonCircuitsandSystemsforVideoTechnology,vol.10,pp.974–979,2000.
[18]I.J.Cox,J.Kilian,F.T.Leighton,andT.Shamoon,“Securespreadspec-trumwatermarkingformultimedia,”IEEETransactionsonImageProcess-ing,vol.6,pp.1673–1687,1997.
[19]I.J.CoxandJ.P.Linnartz,“Somegeneralmethodsfortamperingwithwatermarks,”IEEEJournalonSelectedAreasinCommunications,vol.16,pp.587–593,1998.
[20]W.Zhu,Z.Xiong,andY.Q.Zhang,“Multiresolutionwatermarkingforimagesandvideo,”IEEETransactionsonCircuitsandSystemforVideoTechnology,vol.9,pp.545–550,1999.
[21]M.Barni,F.Bartoilini,V.Cappellini,andA.Piva,“Multiresolutionwa-termarkingforimagesandvideo,”SignalProcessing,vol.66,pp.357–372,1998.
[22]C.I.PodilchukandW.Zeng,“Image-adaptivewatermarkingusingvisualmodels,”IEEEJournalonSelectedAreasinCommunications,vol.16,pp.525–539,1998.
[23]C.H.ChouandY.C.Li,“Aperceptuallytunedsubbandimagecoderbasedonthemeasureofjust-noticeable-distortionprofile,”IEEETransactionsonCircuitsandSystemsforVideoTechnology,vol.5,pp.467–476,1995.
[24] R.C.GonzalezandR.E.Woods,DigitalImageProcessing,2edition,2002.
[25]C.T.HsuandJ.L.Wu,“Dct-basedwatermarkingforvideo,”IEEETrans-actionsonConsumerElectronics,vol.44,pp.206–216,1998.
[26]C.T.HsuandJ.L.Wu,“Multiresolutionwatermarkingfordigitalimages,”IEEETransactionsonCircuitsandSystems,vol.45,pp.1097–1101,1998.
[27]C.T.HsuandJ.L.Wu,“Hiddendigitalwatermarksinimages,”IEEETransactionsonImageProcessing,vol.8,pp.58–68,1999.
[28]C.S.Lu,S.K.Huang,C.J.Sze,andH.Y.M.Liao,“Cocktailwatermarkingfordigitalimageprotection,”IEEETransactionsonMultimedia,vol.2,pp.209–224,2004.
[29]M.O.Rabin,“Digitalizedsignaturesandpublic-keyfunctionsasintractableasfactorization,”TechnicalReportMIT/LCS/TR212,LaboratoryforCom-puterScience,MassachusettsInstituteofTechnology,1979.
[30]G.Z.Wu,Y.J.Wang,andW.H.Hsu,“Robustwatermarkembed-ding/detectionalgorithmforh.264video,”JournalofElectronicImaging,vol.14,pp.13013–0–13013–9,2005.
[31]J.Huang,L.Chen,andY.Q.Shi,“Anintegratedclassifierinclassifiedcoding,”inProceedings,1998IEEEInternationalSymposiumonCircuitsandSystems(ISCAS’98),1998,vol.4,pp.146–149.
[32]T.Y.Chung,M.S.Hong,Y.N.Oh,andS.H.Park,“Digitalwatermarkingforcopyrightprotectionofmpeg2compressedvideo,”IEEETransactionsonConsumerElectronics,vol.44,pp.895–901,1998.
[33]W.Stallings,CryptographyandNetworkSecurity:PrinciplesandPractice,Prentice-Hall,internationalsecondeditionedition,1999.
[34]R.Rivest,A.Shamir,andL.Adleman,“Amethodofobtainingdigitalsignaturesandpublickeycryptosystems,”CommunicationsofACM,1978.
[35]D.Y.Chen,M.Ouhyoung,andJ.L.Wu,“Ashift-resistingpublicwatermarksystemforprotectingimageprocessingsoftware,”IEEETransactionsonConsumerElectronics,vol.6,pp.404–414,2000.
[36]S.D.LinandC.F.Chen,“Arobustdct-basedwatermarkingforcopyrightprotection,”IEEETransactionsonConsumerElectronics,vol.46,pp.415–421,2000.
[37]A.M.Alattar,E.T.Lin,andM.U.Celik,“Watermarkinglowbit-rateadvancedsimpleprofilempeg-4bitstreams,”inProceedings,2003IEEEInternationalConferenceonAcoustics,Speech,andSignalProcess-ing(ICASSP’03),2003,vol.3,pp.513–516.
[38]H.Kii,J.Onishi,andS.Ozawa,“Thedigitalwatermarkingmethodbyusingbothpatchworkanddct,”inProceedings,1999IEEEInternational
ConferenceonMultimediaComputingandSystems,1999,vol.1,pp.895–
899.
[39]B.ChenandG.W.Wornell,“Achievableperformanceofdigitalwater-markingsystems,”inProceedings,1999IEEEInternationalConferenceonMultimediaComputingandSystems,1999,vol.1,pp.13–18.
[40]P.V.K.BorgesandJ.Mayer,“Positionbasedwatermarking,”inProceed-ings,20033rdInternationalSymposiumonImageandSignalProcessingandAnalysis,2003,vol.2,pp.997–1002.
[41]B.ChenandG.Wornell,“Quantizationindexmodulation:Aclassofprov-ablygoodmethodsfordigitalwatermarkingandinformationembedding,”IEEETransactionsonInformationTheory,vol.47,pp.1423–1443,2001.
[42] J.J.EggersandB.Girod,InformedWatermarking,Norwell,2002.
[43]P.H.W.WongandO.C.Au,“Ablindwatermarkingtechniqueinjpegcom-presseddomain,”inProceedings,2002InternationalConferenceonImageProcessing.,2002,vol.3,pp.497–500.
[44]P.H.W.Wong,A.Chang,andO.C.Au,“Onimprovingtheiterativewater-markembeddingtechniqueforjpeg-to-jpegwatermarking,”inProceedings,2004InternationalSymposiumonCircuitsandSystems(ISCAS’04),2004,vol.2,pp.161–164.
[45]P.H.W.WongandO.C.Au,“Acapacityestimationtechniqueforjpeg-to-jpegimagewatermarking,”IEEETransactionsonCircuitsandSystemsforVideoTechnology,vol.13,pp.746–752,2003.
[46]P.H.W.Wong,O.C.Au,andY.M.Yeung,“Novelblindmultiplewater-markingtechniqueforimages,”IEEETransactionsonCircuitsandSystemsforVideoTechnology,vol.13,pp.813–830,2003.
[47]Q.LiandI.J.Cox,“Usingperceptualmodelstoimprovefidelityandprovideinvariancetovalumetricscalingforquantizationindexmodulationwater-marking,”inProceedings,2005IEEEInternationalConferenceonAcoustics,Speech,andSignalProcessing,,2005,vol.2,pp.1–4.
[48]L.M.Marvel,C.G.Boncelet,andC.T.Retter,“Spreadspectrumimagesteganography,”IEEETransactionsonImageProcessing,vol.8,1999.
[49]S.Craver,N.Memon,B.L.Yeo,andM.M.Yeung,“Resolvingrightfulownershipswithinvisiblewatermarkingtechniques:limitations,attacks,andimplications,”IEEEJournalonSelectedAreasinVommunications,vol.16,pp.573–586,1998.