STUDY OF MULTIMEDIA WATERMARKINGTECHNIQUES
Mrs. Chhaya S. Gosavi
Department of Computer EnggCummins College of Engg. For Women,Karvenagar,Pune, IndiaEmail:chhaya.gosavi@cumminscollege.in
Dr. C.S. Warnekar
Department of Information TechnologyCummins College of Engg. For Women,Karvenagar,Pune, IndiaEmail:cswarnekar@hotmail.com
Abstract
—With the recent burgeoning of networked multimediasystems, techniques are needed to prevent illegal copying / forgery in distributed digital audio/ visual/text document. It maybe also desirable to determine where and by how much themultimedia file has been changed from the original due toattacks. This is attributed to increasing instances of hackingduring digital communicationDigital watermarking has been proposed as a solution to theabove problem to protect multimedia document. There are twoimportant issues that watermarking algorithms need to address.Firstly, watermarking schemes are required to providetrustworthy evidence for protecting rightful ownership. Secondly,good watermarking schemes should satisfy the requirement of robustness and resist distortions due to common manipulations(such as truncation, compression etc.)In this paper, various techniques to secure Multimedia data arediscussed.
Keywords-Digital watermarking;DCT;IDCT;DFT; DWT; Singularvalue decomposition; Security.
I.
I
NTRODUCTION
The rapid evolution of the cyber world has greatlyfacilitated the manipulation and transmission of digitaldocuments in text, images, audio, and video forms. Easy accessand replication, however, have led to serious problemsregarding copyright protection and/ or distortion prevention of multimedia documents. Conventionally watermarking is usedfor copyright protection of documents. Presently digitalwatermarking as an offshoot of computer technology haswidened its field of application. Drawing from many relatedfields, such as cryptography, communication theory,information theory, etc., digital watermarking is proving to be apowerful security measure in transmission of multimediadigital documents. Media owners use this technique to insertidentifying information into their document for the purpose of copyright protection. Alternatively they may embed the desiredsignal into another multimedia document for more securedcommunication process.Three qualities are required in digital watermarking:transparency, robustness, and capacity. Transparency refers tothe fact that a watermark embedded image signal closelyresembles its original version. E.g. it is difficult to differentiatebetween an audio signal with watermark and its unmarkedversion. Robustness refers to ability to resist distortion. This istaken care by the invariant properties of the transform.Capacity refers to percentage of watermark signal which maybe embedded in original signal without noticeable distortion inthe quality. However these characteristics are often mutuallycontradictory, so compromises must be made while applyingthem.Most of the existing watermarking algorithms areapplicable to images or video signals. However, the
literatureon intermixing of audio-visual signals to realize watermarkingis comparatively limited. The widespread use of the Internetand the digital audio distribution in MP3 form has made thecopyright protection of digital audio work also more and morenecessary. Some research works have been published on audioto audio watermarking. These approaches work in the timedomain [1], temporal domain [2], DCT domain [3], DWTdomain [4], cepstrum domain [5, 6], or sub band domain [7, 8].In this paper we provide a survey of the latest techniquesthat are employed to watermark images, audio and video. Thepaper is organized in the following sections. In Section 2 wedescribe Image watermarking techniques. In Section 3 weidentify the techniques for audio watermarking. In Section 4 wediscuss the video watermarking techniques. We conclude thispaper in section 5 where we give some guidelines ondeveloping robust watermarking algorithms.II.
I
MAGE
W
ATERMARKING
Basically there are two main types of watermarks that canbe embedded within an image.
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201064http://sites.google.com/site/ijcsis/ISSN 1947-5500
Pseudo-Random Gaussian Sequence
A Gaussian sequence watermark is a sequence of numberscomprising 1 and -1 and which has equal number of 1's and -l'sis termed as a watermark. It is termed as a watermark with zeromean and one variation. Such watermarks are used forobjective detection using a correlation measure.
Binary Image or Grey Scale Image Watermarks
Some watermarking algorithms embed meaningful data inform of a logo image instead of a pseudo-random Gaussiansequence. Such watermarks are termed as binary imagewatermarks or grey scale watermarks. Such watermarks areused for subjective detection.Based on the type of watermark embedded, an appropriatedecoder has to be designed to detect the presence of watermark.If it's a pseudo random Gaussian sequence hypothesis, testingis done to detect the presence of watermark. Suppose W is theoriginal watermark bit sequence and W' is the extractedwatermark bit sequence, then we can calculate bit error rate(BER) to detect the presence of watermark. If the BER is zeroit indicates the presence of watermark; however, if it is one, itindicates absence of watermark. BER is calculated as follows.Suppose D is the retrieved signal and N is the number of bits inwatermark then:……(1)Images can be represented in spatial domain and transformdomain. The transform domain image is represented in terms of its frequencies; however, in spatial domain it is represented bypixels. In simple terms transform domain means the image issegmented into multiple frequency bands. To transfer an imageto its frequency representation we can use several reversibletransform like Discrete Cosine Transform (DCT), DiscreteWavelet Transform (DWT), or Discrete Fourier Transform(DFT). Each of these transforms has its own characteristics andrepresents the image in different ways. Watermarks can beembedded within images by modifying these values, i.e. thepixel values or the transform domain coefficients. Simplewatermarks could be embedded in the spatial domain of imagesby modifying the pixel values or the least significant bit (LSB)values; however, more robust watermarks could be embeddedin the transform domain of images by modifying the transformdomain coefficients. Following Figures shows the result of Spatial domain technique, i.e. LSB modification.
Figure 1. Embedding Watermark (LSB)
Figure 2. Extracting Watermark (LSB)
A.
DCT DOMAIN WATERMARKING
DCT based watermarking techniques are more robustcompared to simple spatial domain watermarking techniques.Such algorithms are robust against simple image processingoperations like low pass filtering, brightness and contrastadjustment, blurring etc. However, they are difficult toimplement and are computationally more expensive. At thesame time they are weak against geometric attacks likerotation, scaling, cropping etc. DCT domain watermarking canbe classified into
Global DCT
watermarking and
Block basedDCT
watermarking. One of the first algorithms presented byCox et al. (1997) used global DCT approach to embed a robustwatermark in the perceptually significant portion of the HumanVisual System (HVS). Embedding in the perceptuallysignificant portion of the image has its own advantages becausemost compression schemes remove the perceptuallyinsignificant portion of the image. In spatial domain itrepresents the LSB however in the frequency domain itrepresents the high frequency components. The main steps of any block based DCT algorithm are as follows:
Steps in DCT Block Based Watermarking Algorithm
1) Segment the image into non-overlapping blocks of 8x82) Apply forward DCT to each of these blocks3) Apply some block selection criteria (e.g. HVS)4) Apply coefficient selection criteria (e.g. highest)5) Embed watermark by modifying the selectedcoefficients.6) Apply inverse DCT transform on each block Most algorithms are classified based on step 3 and 4 i.e. themain difference between most algorithms is that they differeither in the block selection criteria or coefficient selectioncriteria. Based on the perceptual modeling strategyincorporated by the watermarking algorithms they could beclassified as algorithms with:
1)
No Perceptual Modeling:
Such algorithms do not incorporate any perceptual modelingstrategy while embedding a watermark.
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201065http://sites.google.com/site/ijcsis/ISSN 1947-5500
2)
Implicit Perceptual Modeling
Such algorithms incorporate the transform domainproperties for perceptual modeling. The coefficient selectioncriterion is as follows:i) Select those transform coefficients which have largeperceptual capacity, because they allow stronger watermarks tobe embedded and result in least perceptual distortion. DCcomponent satisfy this criteria and hence can be used.ii) Select only those coefficients which are least changed bycommon image processing attacks like low-pass filtering, noiseaddition etc. Low frequency AC components (or highmagnitude coefficients) as well as high magnitude DCcomponents satisfy the above criteria and can be selected.iii) High frequency components are affected by commonimage processing operations hence they are not a good choicefor watermarking.
3)
Explicit Perceptual Modeling
Such algorithms incorporate the HVS properties forperceptual modeling. HVS models allow us to raise or lowerthe strength of the watermark because it takes into account thelocal image properties like contrast, brightness, variance etc.
B.
DWT DOMAIN WATERMARKING
In the last few years wavelet transform has been widelystudied in signal processing in general and image compressionin particular. In some applications wavelet based watermarkingschemes outperforms DCT based approaches.
Characteristics of DWT
i) The wavelet transform decomposes the image into threespatial directions, i.e. horizontal, vertical and diagonal. Hencewavelets reflect the anisotropic properties of HVS moreprecisely.ii) Wavelet Transform is computationally efficient and can beimplemented by using simple filter convolution.iii) Magnitude of DWT coefficients is larger in the lowestbands (LL) at each level of decomposition and is smaller forother bands (HH, LH, HL).iv) The larger the magnitude of the wavelet coefficient themore significant it is.v) Watermark detection at lower resolutions is computationallyeffective because at every successive resolution level there arefew frequency bands involved.vi) High resolution subbands helps to easily locate edge andtextures patterns in an image.
Advantages of DWT over DCT
i) Wavelet transform understands the HVS more closely thanthe DCT.ii) Wavelet coded image is a multi-resolution description of image. Hence an image can be shown at different levels of resolution and can be sequentially processed from lowresolution to high resolution.iii) Visual artifacts introduced by wavelet coded images areless evident compared to DCT because wavelet transformdoesn't decompose the image into blocks for processing. Athigh compression ratios blocking artifacts are noticeable inDCT; however, in wavelet coded images it is much clearer.iv) DFT and DCT are full frame transform, and hence anychange in the transform coefficients affects the entire imageexcept if DCT is implemented using a block based approach.However DWT has spatial frequency locality, which means if signal is embedded it will affect the image locally. Hence awavelet transform provides both frequency and spatialdescription for an image.
Disadvantages of DWT over DCT
Computational complexity of DWT is more compared toDCT. As Feig (1990) pointed out it only takes 54multiplications to compute DCT for a block of 8x8, unlikewavelet calculation depends upon the length of the filter used,which is at least 1 multiplication per coefficient.
DWT Watermarking
DWT based watermarking schemes follow the sameguidelines as DCT based schemes, i.e. the underlying conceptis the same; however, the process to transform the image intoits transform domain varies and hence the resulting coefficientsare different. Wavelet transforms use wavelet filters totransform the image. There are many available filters, althoughthe most commonly used filters for watermarking are HaarWavelet Filter, Daubechies Orthogonal Filters and DaubechiesBi-Orthogonal Filters. Each of these filters decomposes theimage into several frequencies. Single level decompositiongives four frequency representations of the images. These fourrepresentations are called the LL, LH, HL, HH subbands asshown in Fig.3.
Figure 3.
Single level Decomposition using DWT
DWT algorithms can be classified based on their decoderrequirements as Blind Detection or Non-blind Detection. Blinddetection doesn't require the original image for detecting thewatermarks; however, non-blind detection requires the originalimage.
C.
DFT DOMAIN WATERMARKING
DFT domain has been explored by researches because it offersrobustness against geometric attacks like rotation, scaling,cropping, translation etc.LL1 HL1LH1 HH1
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 201066http://sites.google.com/site/ijcsis/ISSN 1947-5500
Search History:
Result 00 of 00
00 results for result for
p.
Study Of Multimedia Watermarking Techniques
With the recent burgeoning of networked multimedia systems, techniques are needed to prevent illegal copying / forgery in distributed digital audio/ visual/text document. It may be also desirable to determine where and by how much the…
(More)
1.4K
Reads
11
Readcasts
3
Embed Views
More From This User
Notes
Load more
