A Video Watermarking Scheme Based on theDual-Tree Complex Wavelet Transform

TERMPAPER REVIEW

guided by: presentation by:

Smt. T.Geetamma A.Srinivasa Rao

Assistant professor 12341A0402

dept. of ECE ECE

GMRIT GMRIT

ABSTRACT

In this method, the video is watermarked in such a way that its display isnot permitted if a compliant video player detects the watermark. A watermark that isrobust to geometric distortions (rotation, scaling, cropping) and lossy compression isrequired in order to block access to media content that has been re-recorded with acamera inside a movie theater.

This paper proposed a new video watermarking algorithm for playbackcontrol that takes advantage of the properties of the dual-tree complex wavelettransform. This transform offers the advantages of the regular and the complexwavelets. This method relies on these characteristics to create a watermark that isrobust to geometric distortions and lossy compression. The proposed scheme issimple to implement and outperforms comparable methods when tested againstgeometric distortions.

INTRODUCTION

• PIRACY means the practice of selling, acquiring,copying or distributing copyrighted material withoutpermission.

• Although digital technology has brought many benefitsto the content creators and the public, it has alsoincreased the ease by which movies can be pirated.

• In this paper the advanced method called videowatermarking using dual tree complex wavelettransform was proposed in order to protect thecopyrights.

WHAT IS WATERMARKING??

• Hiding a message signal into a host signal, without any perceptualdistortion of the host signal.

• The main application of digital watermarking is in copyrightprotection.

• Type of watermarks: 1.visible 2.invisible

LITERACY SURVEY

Ref1:Persistent access control to prevent piracy of digital information[1997]

1. Watermarking technology was first introduced in 1990’s whereplayback control application, the watermark embedded in the videosequence.

2. With help of this playback control application it will provideinformation that whether video players are authorized to display thecontent or not.

Ref2:Rotation, scale and translation invariant digital image watermarking[1999]

• In this paper author proposed an image watermarking method based on the Fourier–Mellin transform is proposed

• The embedded marks may be designed to be unaffected by any combination of rotation, scale and translation transformations.

• The scheme is robust to rotation and scaling but weak to distortions caused by lossy compression.

Ref3:Rotation, scale, and translation resilient watermarking for images[2001]

1. The watermark is embedded into a 1-D signal, which is obtainedby taking the Fourier transform of the image, resampling it in tolog-polar coordinates, and integrating along the radial dimension.

2. This method is robust to rotation, scaling, and translation and lossycompression which is one of advantage compared to previousresults.

3. However this scheme cannot withstand cropping.

Ref4: DWT based high-capacity blind video watermarking ,in variant to geometrical attacks[2003]

1. In this two watermarks are employed. The first one is used to embedthe message while the second one, a 0-b watermark, is employed asa geometric reference.

2. Once the reference watermark has been changed, the decoderassumes that there is no watermark embedded in the content and,therefore, does not search for the hidden message.

3. Even though this method is tricky secure, introducing twowatermarks in the same video is complex and may effect the hostsignal.

PROPOSED METHOD

The proposed method has following steps to create DTCWT algorithm:

A. Creating the Watermark.

B. Embedding the Watermark. 1) Perceptual Masks

2) Adding the Watermark

C. Decoding the Watermark.

Brief Introduction to the DT CWT

• The Dual-tree complex wavelet transform(DTCWT) calculates the complex transform of asignal using two separate DWT decompositions.

• This transform has the desirable properties of theDWT and the CWT such as perfect reconstruction,approximate shift invariance, good directionalselectivity.

• The main difference being DWT &DTCWT is that ituses two filter trees instead of one.

• The watermark is a random set of 1’s and -1’s. Aone-level DTCWT is applied to this watermark andthe coefficients of this transformation become thedata that are embedded into the video sequence.Every frame of the original video sequence istransformed with a four-level DT CWT

• The dual-tree approach provides wavelet coefficients that are

approximately shift invariant i.e. small shifts in the input signal

will not cause major variations in the distribution of energy of DT

CWT coefficients at different scales.

• The typical response of the high-pass decimation filter for each

filter tree is as shown in figure. the filters used in tree B are

designed to produce outputs at sample locations that are discarded

in tree A.

• For each level, there are six sub bands that correspond to the outputof six directional filters oriented at angles of ±15°, ±45°, ±75°,

Typical impulse response of HPF

Creating the Watermark

• In this method, the watermark is inserted in every frame of

the video sequence.

• The watermark is a 2-D array that is 64 times smaller than

the video frame where it will be embedded.

• The use of the same K for β consecutive frames offers some

robustness to temporal synchronization attacks. This is as

long as β is small enough (so that an attacker cannot detect

and remove the watermark by frame averaging) but long

enough (so that if some frames are dropped, the watermark

can still be detected).

• DTCWT is a redundant transformation. Thus, somecomponents of the arbitrary pseudorandom sequence in theDT CWT domain may be lost during the DT CWT inversetransformation process.

𝑲𝒂 :constant

𝑲𝒇 :positive integer number

that changes every β frames.

𝐾 = 𝐾𝑎 +𝐾𝑓

Embedding the WatermarkEmbedding the watermark carried out in 2 steps i. Perceptual Masks

ii. Adding the Watermark

Creating perceptual masks:

• Containing watermark in a frame might significantly decrease the content’s fidelity since the human visual system is very susceptible to changes in the low frequencies.

• This can be overcome by using masks which hide the watermark to be visible to human eye.

𝒎𝒂𝒔𝒌𝟑,𝒅 =↓𝟐 𝑭𝑯𝟐,𝒅 ∗𝒉𝑳𝑷

∆for d=1,2,…,6

where ℎ𝐿𝑃=

1

4

1

41

4

1

4

and magnitude of level2 𝐹𝐻2,𝑑 =

𝐹𝐻2,𝑑(0,0) 𝐹𝐻2,𝑑(0,𝑀

4− 1)

. .

𝐹𝐻2,𝑑(𝑁

4− 1,0) 𝐹𝐻2,𝑑(

𝑁

4− 1,

𝑀

4− 1)

• The masks for level 4 subbands are created in similar way. 𝑚𝑎𝑠𝑘4,𝑑 for d=1,2,…..,6.

Adding the watermark:

• For each frame, the watermark’s complex high-frequency coefficients 𝑊𝐻1……𝑊𝐻6 are added to the magnitudes of the coefficients of level 3 and level 4.(𝐹𝐻3,1…… 𝐹𝐻3,6 respectively)

𝑭𝑾𝟑,𝒅 = 𝑭𝑯𝟑,𝒅 +∝ 𝒎𝒂𝒔𝒌𝟑,𝒅𝑾𝑯𝒅 𝑾𝑯𝒅

𝑾𝑯𝒅 𝑾𝑯𝒅

for d=1,2,….,6 and ∝ is scalar factor

where 𝐹𝐻3,𝑑 =

𝐹𝐻3,𝑑(0,0) 𝐹𝐻3,𝑑(0,𝑀

8− 1)

. .

𝐹𝐻3,𝑑(𝑁

8− 1,0) 𝐹𝐻3,𝑑(

𝑁

8− 1,

𝑀

8− 1)

and is 2-D array formed with the phase of the complex elements 𝑭𝑾𝟑,𝒅

• Once 𝑭𝑾𝟑,𝒅 and 𝑭𝑾𝟒,𝒅 and are obtained, they replace 𝑭𝑯𝟑,𝒅 and 𝑭𝑯𝟑,𝒅 when computing the inverse DT CWT that provides the watermarked frame.

Decoding the Watermark

• Embedding and adding the watermark to a video frames is not the only work of a

creator but having the knowledge of decoding the watermark also plays an

important role.

• The decoding process is blind, that is, the watermark is decoded without relying

on any information from the original video file.

• Essentially, the decoder performs the inverse operations of the encoder.

• The masks for levels 3 and 4 are obtained via 𝒎𝒂𝒔𝒌𝟑,𝒅 and 𝒎𝒂𝒔𝒌𝟒,𝒅.

• The arrays 𝑖𝑚𝑎𝑠𝑘3,1…, 𝑖𝑚𝑎𝑠𝑘3,6 and 𝑖𝑚𝑎𝑠𝑘4,1…. 𝑖𝑚𝑎𝑠𝑘4,6 are obtained infollowing way:

𝑖𝑚𝑎𝑠𝑘𝑠,𝑑 =

1

𝑚𝑎𝑠𝑘𝑠,𝑑 0,0… . .

1

𝑚𝑎𝑠𝑘𝑠,𝑑 0,𝑀

2𝑆−1

: … . . :1

𝑚𝑎𝑠𝑘𝑠,𝑑𝑁

2𝑆−1,0

… . .1

𝑚𝑎𝑠𝑘𝑠,𝑑𝑁

2𝑆−1,

𝑀

2𝑆−1

for s=3,4 &d=1,2,..,6

• The watermarked level 3 and level 4 coefficients 𝑭𝑾𝟑,𝒅 and 𝑭𝑾𝟒,𝒅are multiplied by theimask arrays in order to compensate for the different weights associated with every coefficient during the watermark embedding process.

𝑭′𝑾𝒔,𝒅=𝑭𝑾𝑺,𝒅 𝒊𝒎𝒂𝒔𝒌𝒔,𝒅 for s=3,4 and d=1,2,….,6.

• Next, W’ the level-1 DT CWT representation of the decoded watermark w’ , is obtained.

• The six sub bands with details 𝑾′𝑯𝟏……𝑾′𝑯𝟔 can be estimated.

EXPERIMENTAL RESULTS• In order to study the performance of this method, DTCWT is compared with results

against two algorithms that employ the regular DWT.

• The first method we use as reference is basically the same algorithm as proposed in thispaper except DWT replaces DT CWT. We will refer to this method as DWT1.

• The second method is the one presented in ref9 which is also based on DWT. In thismethod, which we denote as DWT2.

• To tested the robustness of this method to common distortions. In one experiment,watermarks were decoded after the video sequences had gone through some scaling andcropping distortions.

• For the second test, the video sequences were rotated by a few degrees and thewatermark was later decoded. Here the effects of lossy compression are also tested

• Finally all of these distortions: scaling, rotation, cropping, and lossy compression wereput together as a joint attack.

A. Frame Scaling and Cropping

• Every video sequence was scaled up by 5%, 10%, and 15% usingbicubic interpolation.

• The frames were later cropped to fit their original size (176X144).

• From these results, we notice that DT CWT is able to withstand ascaling and cropping attack, particularly for scales of 5% and10%. DWT2, however, performs better than the other schemes forthis type of attack.

• A visual example of this process can be seen in Fig. Watermarkedframe of the sequence Suzie is scaled and then cropped (a) 5%, (b)10%, and (c) 15% scaling.

B. Frame Rotation

• Robustness to frame rotation was then tested. Each frame wasrotated counter clockwise by 3°,6°,and 9°.

• Bilinear interpolation was employed and the resulting imageswere cropped to fit the QCIF format.

• It can be observed that DTCWT is more robust to rotation thanthe other two methods.

• Although DTW1 is able to decode 100% of the watermarkswhen the frames are rotated by 3° , the scheme can only recover30% of the watermarks once rotation has increased to 6° .DWT2 offers very poor performance for this particular type ofdistortion.

C. Compression

• In order to test the robustness of the proposed scheme tocompression, the video sequences is encoded usingH.264/AVC. Every 15th frame was set to be an I-frame and therest were chosen to be P-frames.

• The quantization parameter QP for both frames was set to 15,which results in a compression ratio of around 40 : 1.

• In this instance, the three watermarking methods demonstratedrobustness to compression since all of the watermarks weredecoded.

• an example of a compressed frame can be seen in Fig.

D. Joint Attack

• The final experiment involved all of the previous attackstogether.

• For this joint attack, we scaled the video frames by 5%and rotated them by 5° . The frames were later cropped tofit their original size (176X144) and H.264/AVC wasused to compress the video sequences.

• Results for the DT CWT indicate that the method cansuccessfully survive a joint attack. 92% of thewatermarks were detected even though the videosequences had gone through scaling, rotation, cropping,and compression.

COMPARISON OF NORMALIZED CORRELATION VALUES FOR THREE WATERMARKING METHODS: DT CWT, DWT1, AND DWT2.

WATERMARKED SEQUENCES ARE SUBJECTED TO SCALING (BY 5%, 10%, AND 15%), ROTATION (BY 3° , 6° and 9° ), CROPPING,

H.264 COMPRESSION WITH A QF=15, AND A JOINT ATTACK THAT INVOLVES SCALING (BY 5%), ROTATION (BY 6° ), CROPPING,

AND COMPRESSION.

CONCLUSION• DT CWT provides important features, such as perfect reconstruction, shift invariance, and

good directional selectivity.

• The robustness of our method was tested against several attacks, which included lossy

compression, rotation, scaling, cropping, and a joint attack.

• The joint attack was employed to simulate a video sequence that has been recorded from a

movie screen with a handheld camcorder and then stored in a digital form. Our method

successfully detected the presence of the watermarks in 92% of the corrupted video

sequences.

• DTCWT method is simple to implement this is important when considering the additional

cost and complexity to DVD players. All of these characteristics make this algorithm

suitable for the playback control of digital video.

REFERENCES:• A Video Watermarking Scheme Based on the Dual-Tree Complex Wavelet Transform. Lino E. Coria, Member,

IEEE, Mark R. Pickering, Member, IEEE, Panos Nasiopoulos, Member, IEEE, and Rabab Kreidieh Ward, Fellow,IEEE VOL. 3, NO. 3, SEPTEMBER 2008.

• P. B. Schneck, “Persistent access control to prevent piracy of digital information,” Proc. IEEE, vol. 87, no. 7, pp.1239–1249, Jul. 1999.

• I. J. Cox, M. L. Miller, and J. A. Bloom, Digital Watermarking. San Francisco, CA: Morgan Kaufmann, 2002.

• J. J. K. O’Ruanaidh and T. Pun, “Rotation, scale and translation invariant digital image watermarking,” in Proc. Int.Conf. Image Processing,1997, pp. 536–539.

• C.-Y. Lin, M.Wu, J. A. Bloom, I. J. Cox, M. L. Miller, and Y. M. Lui, “Rotation, scale, and translation resilientwatermarking for images,” IEEE Trans. Image Process., vol. 10, no. 5, pp. 767–782, May 2001.

• C. V. Serdean, M. A. Ambroze, M. Tomlinson, and J. G.Wade, “DWTbased high-capacity blind videowatermarking, invariant to geometrical attacks,” Proc. Inst. Elect. Eng., Vis., Image Signal Process., vol. 150,pp.51–58, Feb. 2003.

• P. Bas, J. M. Chassery, and B. Macq, “Geometrically invariant watermarking using feature points,” IEEE Trans.Image Process., vol. 11, no. 9, pp. 1014–1028, Sep. 2002.

• P. W. Chan, M. R. Lyu, and R. T. Chin, “A novel scheme for hybrid

• digital video watermarking: Approach, evaluation and experimentation,” IEEE Trans. Circuits Syst. Video Technol.,vol. 15, no. 12, pp.1638–1649, Dec. 2005.