Content uploaded by Deepa Kundur
Author content
All content in this area was uploaded by Deepa Kundur on Sep 05, 2017
Content may be subject to copyright.
A preview of the PDF is not available
Multiresolution Digital Watermarking: Algorithms and Implications for Multimedia Signals
Abstract
y the Natural Sciences and Engineering Research Council (NSERC) of Canada and by the Communications and Information Technology Ontario (CITO). approach, based on channel estimation theory, only requires the use of a key for watermark extraction. Each technique demonstrates improved robustness against signal distortions such as cropping, filtering, perceptual coding, and rescaling, in comparison to previously proposed methods of its class. Simulations verify our theoretical observations and demonstrate the feasibility of digital watermarking for use in multimedia standards. I would like to sincerely thank my supervisor, Professor Dimitrios Hatzinakos, for bringing the problem of digital watermarking to my attention, for suggesting the application of data fusion, and for his feedback during the development of this work. His confidence in my abilities as a researcher is greatly appreciated. I wish to also acknowledge the members of my Ph.D. thesis committees, Professors Gonzalo R. Arc
... Uşurinţa cu care se pot face copii identice face pirateria o problemă majoră în securitatea informaţională. Se prezintă trei situaţii considerate în aplicaţiile obişnuite de securitate şi deficienţele tehnicilor de securitate în protejarea produselor multimedia [Kun99]. ...
... În cele ce urmează, se enumeră câteva domenii care au contact cu marcarea transparentă [Kun99]. ...
... Decodarea se face prin corelaţie normalizată între estimarea marcajului şi marcajul original. Au fost propuse mai multe scheme folosind transformata wavelet [IMK99,Kun99,XBA97,ZXZ98], diferenţa fiind modul de ponderare pentru a diminua efectele vizibile. Majoritatea înserează marcajul în subbenzi de detaliu, cu mici excepţii [JSSK02]. ...
English: Contributions to digital watermaring of still images in the wavelet transform
Corina Naforniţă
Scientific advisors: Prof.Dr.Eng. Monica Borda (UTC-N),
Prof.Dr.Eng. Alexandru Isar (UPT)
The thesis make an analysis of copyright protection schemes design for still images using watermarking in the wavelet domain, having 226 pages, 6 chapters, bibliography and a list of my own publications. The first three chapters define the watermarking techniques, review the main results already reported in literature. The chapters 4 and 5 present my original contributions being proposed several architectures of watermarking systems. All the results have already been published in the Proceedings of prestigious national or international conferences. Some af them have been cited by others authors working in the same field. The chapter 6 contains the conclusions and personal contributions. These are refering to: the classification and critical appreciation of the watermarking methods and attacks, the conclusion being that the digital watermaring in the wavelet transform domain is the most promising; where should the watermark be inserted (in the spatial or transform domain) and to to obtain a robust watermark in what resolution levels of DWT (all, low or high resolution) and in what subbands; what detector architecture ensure the best estimete of watermark; how to make perceptual watermarks; how to evaluate robustness of perceptual watermarks. These contributions are:
1. The informed watermarking [NI03] in the diagonal detail subband of the first resolution level selecting the coefficients to be watermarked using experimental tresholds. 2. The informed watermarking [NBK04] in the all resolution levels, using the same method to select the wavelet coefficients. Detection is made from all resolution levels, or from the last resolution level, less affected by common signal processing. 3. A comparison of our method [NBK04] with a quantization based watermarking method – Kundur and Hatzinakos – in the wavelet domain, or the method in the DCT domain – Cox et al. [CKLS97]. In the next proposed method, [NB05], the images watermarked have a lower watermark strength, hence the quality of the images are comparable, our method being superior to the method proposed by Cox et al. 4. For detecting the watermark, we analyze three types of detectors; the third detector, max-correlation, extracts every estimate of the watermark, [Naf05b], and the watermark with highest correlation with the original one is the response of the detector. 5. The third watermarking method with non-blind detector, selects the wavelet coefficients on the basis of a statistical analysis, [NIB05]. As such, fewer coefficients are affected and the watermark is more robust. 6. A new type of perceptual mask is proposed [NIB06a], using a perceptual mask: data can be embedded in all detail wavelet coefficients, with variable watermark energy. The method is more robust while the quality of the watermarked image is superior. 7. An extension of this method is proposed, [NIB06b], which hides the watermark in lower frequency subbands. This makes the watermark more attack resilient. 8. The evaluation of robustness for perceptual watermarks only when invisibility criteria are satisfied, [Naf07a]. 9. The most important contribution is the method [Naf07b], the watermark is inserted in all resolution levels; the robustness of this method is superior to the one presented by Barni et al. and the attack is resilient to the possible erasure of the first level detail subbands. The third method uses three types of detection i) from all resolution levels, ii) separately from each level, max-level, considering the maximum detector response from each level and iii) separately from each subband, max-subband, considering the maximum detector response from each subband. The method is the most attack resilient comparing with [BBP01], given the fact that the insertion is made on all subbands and on all resolution levels and we use the diversity of the wavelet transform in the detection process. The three detectors can be combined to optimize the watermark extracting, being the final result of the thesis.
Romanian: Contribuţii la marcarea transparentă a imaginilor în domeniul transformatei wavelet
Corina Naforniţă
Conducători şt.: Prof.Dr.Ing. Monica Borda (UTC-N), Prof.Dr.Ing. Alexandru Isar (UPT)
Lucrarea analizează construcţia sistemelor de protecţie a drepturilor de autor, la imagini, prin marcare transparentă (watermarking) în domeniul transformatei wavelet, are 226 de pagini, 6 capitole, lista bibliografică, lista publicaţiilor personale. În primele capitole e analizată literatura şi se definesc tehnicile de marcare. Capitolele 4 şi 5 prezintă contribuţiile originale, fiind propuse mai multe arhitecturi de sisteme de marcare transparentă, rezultatele fiind publicate în volumele unor conferinţe internationale prestigioase, din ţară sau străinătate. Unele dintre aceste articole au fost citate de alţi cercetători din domeniu. În capitolul 6 sunt prezentate concluziile şi contribuţiile personale. Ele se referă la: clasificarea şi aprecierea critică a metodelor de marcare transparentă şi a atacurilor, subliniind că marcarea în domeniul transformatei wavelet, mai ales cea adaptivă, este promiţătoare; la modul de plasare al marcajului pentru creşterea robusteţii lui: în domeniul spaţial sau al unei transformate, în care nivel de descompunere al DWT (toate, cele de rezoluţie inferioară sau superioară), în care din subbenzile DWT; la arhitectura de detector ce asigură extragerea optimă a marcajului; la cum să se facă marcarea perceptuală; la cum se evaluează robusteţea marcajelor perceptuale. Contribuţiile originale (cap. 4 şi 5) sunt:
1. Marcare informată [NI03] în subbanda de detalii diagonale, primul nivel de rezoluţie, sau în toate subbenzile primului nivel, folosind praguri experimentale pentru selectarea coeficienţilor care se marchează. 2. Marcare informată în cele trei nivele de rezoluţie, cu aceeaşi selecţie a coeficienţilor wavelet. Detecţia se face din toate nivelele, sau din ultimul nivel de rezoluţie, mai puţin afectat de prelucrări de semnal. 3. Comparaţie a metodei propuse [NBK04], cu cea prezentată de Kundur şi Hatzinakos de tip cuantizare în domeniul wavelet, respectiv faţă de cea din domeniul DCT de tip spread-spectrum prezentată de Cox et al. Apoi [NB05], intensitatea de marcare este redusă faţă de [CKLS97], imaginile nu sunt vizibil afectate de procesul de marcare, calitatea imaginilor este comparabilă, metoda fiind superioară. 4. Pentru detecţia marcajului la a doua metodă am analizat trei tipuri de detectoare; al treilea detector, max-correlation, estimând marcajul cu corelaţie maximă cu cel original [Naf05b]. 5. Am propus [NIB05] o abordare statistică pentru a selecta mai puţini coeficienţi wavelet în care se înserează marcajul, dar răspunsul detectorului este mai bun, coeficienţii mari fiind mai robuşti. Selecţia pragurilor e bazată pe proprietăţile statistice ale coeficienţilor wavelet. 6. Am propus un nou tip de mascare perceptuală [NIB06a] ce ascunde datele în toţi coeficienţii wavelet de detaliu, cu intensitate de marcare variabilă, cu o mască perceptuală; metoda este mai robustă iar calitatea imaginilor marcate este superioară faţă de [BBP01].7. O extensie a metodei [NIB06b] ascunde imperceptibil marcajul şi în subbenzile de joasă frecvenţă, ce creşte robusteţea. 8. Evaluarea robusteţii marcajelor perceptuale doar când criteriile de invizibilitate sunt satisfăcute [Naf07a]. 9. Cea mai importantă contribuţie este metoda [Naf07b], cu marcaj în toate nivelele, util în cazul ştergerii subbenzilor de frecvenţă înaltă care conţin marcajul la [BBP01]. Am propus şi evaluat trei detectoare: i) din toate nivelele de rezoluţie, ii) separat din fiecare nivel, max-level, şi iii) separat din fiecare subbandă, max-subband. Metoda este mai bună la diverse atacuri decât cea din [BBP01], fapt explicat prin estimatul mai bun al măştii perceptuale şi diversitatea transformării wavelet. Rezultatele celor trei detectoare pot fi fuzionate, ameliorându-se în continuare extragerea marcajului, fiind rezultatul final al tezei de doctorat.
... Yan and Guo [10] converted the LP coefficients to reflection coefficients, which were then transformed to inverse sine (IS) parameters. Watermark embedding was achieved by modifying the IS parameters using odd-even modulation [11]. ...
... The header of each package consists of a 15-bit message produced by a [11,15] BCH encoder [39]. The message contains information in two parts: 7 bits indicating the allocated position and 4 bits specifying the total length. ...
... Step 4 Partition the watermark bit sequence into packages in accordance with the size of the embeddable segment. The location and size of each watermark package are saved as a 15-bit message using a (15,11) BCH encoder, and the resulting BCH code is combined with scrambled watermark bits to form a package. ...
This paper outlines a package synchronization scheme for blind speech watermarking in the discrete wavelet transform (DWT) domain. Following two-level DWT decomposition, watermark bits and synchronization codes are embedded within selected frames in the second-level approximation and detail subbands, respectively. The embedded synchronization code is used for frame alignment and as a location indicator. Tagging voice active frames with sufficient intensity makes it possible to avoid ineffective watermarking during the silence segments commonly associated with speech utterances. We introduce a novel method referred to as adaptive mean modulation (AMM) to perform binary embedding of packaged information. The quantization steps used in mean modulation are recursively derived from previous DWT coefficients. The proposed formulation allows for the direct assignment of embedding strength. Experiment results show that the proposed DWT-AMM is able to preserve speech quality at a level comparable to that of two other DWT-based methods, which also operate at a payload capacity of 200 bits per second. DWT-AMM exhibits superior robustness in terms of bit error rates, as long as the recovery of adaptive quantization steps is secured.
... The binary watermark is also randomly generated using a uniform distribution and is set to be of the same length as ckey(i). The watermark bit f(i) is embedded into the coefficient ckey(i) through an appropriate quantization procedure [18,20]. In the final stage, the corresponding Lth level inverse wavelet transform of the marked image components is computed to form the tamper-proofed image. ...
... Watermark extraction is performed by taking Lth level discrete wavelet transform (DWT) of the given image and the coefficient selection key ckey(i) is used to determine the marked coefficients. A quantization function Q(.) is applied to each of these coefficients to extract the watermark values f(i) [18]. Thus, for authentication, the authorized user's public key ckey(i) is applied to the extracted watermark to obtain the identification code f(i). ...
In this paper a brief overview of digital watermarking is given and a wavelet based fragile image watermarking scheme is proposed. For detection of tampering in an image, a Tamper Assessment Function (TAF) is evaluated and compared to a threshold. The watermark is embedded at different level of wavelet decomposition and attacks based on average filtering and JPEG compression with different compression ratios have been tested. The results show that the tampering can be detected by choosing the threshold suitably.
... Cox's spread spectrum watermarking [14] and Kundur's multi-resolution watermarking [15]. Figure A.1 shows the adopted communications channel model used for generic digital watermarking systems. ...
... Correction Code (ECC), which introduces sufficient data redundancy as a part the payload. The proposed ECC, claimed to correct most of the random errors generated during embedding process, is BCH (63,7,15). For some images, error bits may be concentrated in particular areas of the image and considered as bursts of errors, which leads to too many error bits in one codeword; in such cases, ECC is of no use to recover the data. ...
... An authentication scheme that does not hinder the normal use of the speech signal is needed. Fragile watermarking has provided us such a mechanism [26,15,28]. A secure mark is generated first, which may be independent of the host speech or generated from the speech features. ...
... The modification to the IS parameters are based on odd-even modulation [26] ...
This paper proposes an improved semi-fragile speech watermarking scheme by quantization of linear prediction (LP) parameters, i.e., the inverse sine (IS) parameters. The spectral distortion due to watermark embedding is controlled to meet the ‘transparency’ criterion in speech coding. A modified bit allocation algorithm combined with watermarking is developed to determine the quantization step so that the ‘transparency’ requirement is satisfied. Due to the statistical nature, the LP coefficients estimated from the watermarked speech signal are different from the watermarked LP coefficients even in the absence of attacks. This effect is the cause of increase in decoding error and minimum authentication length. To tackle this problem, an Analysis by Synthesis (AbS) scheme is developed to reduce the difference between the estimated LP coefficients and the watermarked ones. The watermark detection threshold and minimum authentication length are then derived according to the probability of error and the signal to noise ratio (SNR) requirements. Experimental results show that the proposed AbS based method can effectively reduce the difference between the watermarked IS parameter and the extracted IS parameter when there is no attacks. In addition, the modified bit allocation algorithm can automatically find the appropriate quantization step used in the odd-even modulation so that the transparency requirement is satisfied.
... Chen and Zhu [10] inserted watermark bits inside codebook indices corresponding to the LP coefficients obtained from multistage vector quantization (MSVQ). Yan and Guo [11] converted LP coefficients into inverse sine (IS) parameters and then manipulated the IS parameters via odd-even modulation [12]. ...
Watermarking is an important measure for protecting proprietary digital multimedia data. This paper presents a novel approach to achieving robust and imperceptible speech watermarking on a frame-by-frame basis. The proposed method employs two modules operating in the fast Fourier transform (FFT) domain. The first module is referred to as downward progress quantization index modulation. It modulates the vector norms drawn from FFT coefficients according to a guideline deduced from human auditory masking properties. The second module is referred to as boundary-constrained iterative adjustment. It provides a smooth transition across frames in the resulting speech waveform. Experiment results confirm the imperceptibility of the proposed modulation scheme in terms of the mean opinion score of listening quality objective (MOS-LQO) based on the perceptual evaluation of speech quality (PESQ) metric. The proposed watermarking method matched and exceeded the performance of five state-of-the-art methods in terms of robustness against common speech processing attacks.
... We have seen the watermarking of some visual marks in currency notes and papers [16] which is currently extended in multimedia data [9] using digital technology. There are many applications of watermarking. ...
... It is also of interest in non-security applications such as in commerce and in focus, it has gained interest in applications involving captioning, maintenance of audit trails, and embedding hyperlink Information in "hyper-media.", temper assessment/authentications, cover communication and trace applications etc [5][6]. ...
In this paper we propose a transparent, computationally simple and fast spatial domain data hiding technique for content (image and audio) authentication purposes. It can also be used in data captioning, copyright protection, confidentiality and integrity over the property distributed in digital formats over various transmission channels but its fragile nature prevents it to deliver efficient performance in such robust applications. Our method is based on concept that an image (host) can be transformed to another known image (target) by embedding controlled amount of distortion at that particular pixel value controlled through payload bits. Results obtained using MATLAB show that for lower distortion levels, PSNR values as high as 53dB while for appreciable distortions, not below the 35dB level are easily achievable. For our case, the capacity is image and distortion control parameter dependent, hence ranges from 500 bits to 33Kb for a 512x512x8 image requiring a computation time of 105 ms. Moreover, the same technique has also been applied to audio data and even better results obtained are reported. The attractive feature of the proposed algorithm is its simplicity in both embedding and detection of the payload bits, computational efficiency and apparent nature of marked information.
... To establish a more quantitative measure of the added noise to an image, the peak signal-to-noise ratio (PSNR) is used. The PSNR is defined as [14]: ...
In this paper, a proposed face recognition technique for still images is developed and evaluated. It employs wavelet decomposition as a preprocessing step. Then, Principal Component Analysis (PCA) approach is applied on the second level approximation subband to extract the main features of the face. Finally, Probabilistic Radial Basis Function Neural Network (PRBF NN) is used for the image classification. This technique gives better performance in terms of accuracy than published techniques applied on the ORL face database (more than 96% average accuracy). In addition, it possesses high robustness against partial occlusion and added noise.
ResearchGate has not been able to resolve any references for this publication.