#### DMCA

## The Watermark Copy Attack (2000)

Citations: | 74 - 11 self |

### Citations

349 | Multimedia watermarking techniques,”
- Hartung, Kutter
- 1999
(Show Context)
Citation Context ...ormation on the statistics of the image and prior information about the statistics of the noise/watermark, we can use a ML-estimator. The estimator is given by: ^ x = arg max ~ x2R N fln pw (yj~x)g ; =-=(3)-=- where pw (:) is the probability density function of the watermark. The ML-estimate has a closed form solution for the two cases when the watermark has either a Gaussian or a Laplacian distribution. I... |

217 | Analysis of multiresolution image denoising schemes using generalized Gaussian and complexity priors,”
- Moulin, Liu
- 1999
(Show Context)
Citation Context ...e following minimization problem: ^ x k+1 = arg min ~ x2R N f 1 2 2 n ky ~ x k k 2 + w k+1 kr k k 2 g; (8) where w k+1 = 1 r k 0 (r k ); (9) r k = x k x k k x ; (10) 0 (r) =s[( )]sr jrj 2s; (11) where k is the number of iterations, andsis again the shape parameter of the generalized Gaussian distribution. In this case, the penalty function is quadratic for asxed weighting function w. Assumin... |

144 | A Fair Benchmark for Image Watermarking Systems”,
- Kutter, Petitcolas
- 1999
(Show Context)
Citation Context ...nformation about the statistics of both the cover image and the watermark/noise, then we can use an MAP-estimator. The MAP-estimator is given by: ^ x = arg max ~ x2R N fln pw (yj~x) + ln p x (~x))g ; =-=(6)-=- where p x (:) is the probability density function of the cover image. It is not possible tosnd a closed form solution for the MAP-estimate in all cases. We model the image as stationary generalized G... |

96 | Some general methods for tampering with watermarks. Selected Areas in Communications,
- Cox, Linnartz
- 1998
(Show Context)
Citation Context ... of the denoising theory. Independent of the watermarking technology employed, we can model the watermarking process as an addition of a watermark to the cover image in the spatial domain: y = x + w; =-=(1)-=- where y is the stego image, x the original/cover image and w the watermark. Considering the stego image as a noisy image, then the watermark is the noise and we can compute an estimate of the noise/w... |

95 | A stochastic approach to content adaptive digital image watermarking”, in
- Voloshynovskiy
- 1999
(Show Context)
Citation Context ...the shrinkage solution of image denoising problem previously used only in the wavelet domain can easily be obtained from Equation 8 in the next closed form: ^ x = x + max(0; jy xj T )sign(y x); (14) where T = 2 n x p 2 is the threshold for practically important case of Laplacian image prior. This coincides with the soft-thresholding solution of the image denoising problem. 11 5. COPY INSERTIO... |

65 | Public watermarks and resistance to tampering,” in
- Cox, Linnartz
- 1997
(Show Context)
Citation Context ... image, then the watermark is the noise and we can compute an estimate of the noise/watermark ^ w by taking the dierence between the estimate ^ x of the cover image and the stego image: ^ w = y ^ x: (=-=2)-=- 3 Estimator w x y Figure 3. Watermark prediction through denoising. This approach of predicting the embedded watermark through denoising is illustrated in Figure 3. If assume to have no prior informa... |

46 |
Probability and Stochastic Processes for Engineers.
- Helstrom
- 1991
(Show Context)
Citation Context ... the watermark has either a Gaussian or a Laplacian distribution. If the watermark has a Gaussian distribution the ML-estimate is given by the local mean of y: ^ x = localmean(y); Gaussian watermark: =-=(4)-=- On th other hand, if the watermark features a Laplacian distribution the solution of the ML-estimate is given by the local median: ^ x = localmedian(y); Laplacian watermark: (5) Computing the ML-esti... |

39 | Generalized watermarking attack based on watermark estimation and perceptual remodulation
- Voloshynovskiy, Pereira, et al.
- 2000
(Show Context)
Citation Context ... phenomena of the HVS. 7,12 In this work we decided to used the noise visibility function (NVF) proposed by Voloshynovskiy et al. 14 The noise visibility function is dened as: NV F = ! ! + 2 x ; (15) where 2 x the cover image variance, and ! is a local weighting function dened by: ! =s[( )]s1 jrj 2s; (16) where r = x x x . is a tuning parameter dened by: = 100 2 max (17) There NVF c... |

29 |
Removing spatial spread spectrum watermarks.
- Langelaar, Lagendijk, et al.
- 1998
(Show Context)
Citation Context ..., which allows us to derive an iterative optimal solution. 14 Then equation (6) is reduced to the following minimization problem: ^ x k+1 = arg min ~ x2R N f 1 2 2 n ky ~ x k k 2 + w k+1 kr k k 2 g; (8) where w k+1 = 1 r k 0 (r k ); (9) r k = x k x k k x ; (10) 0 (r) =s[( )]sr jrj 2s; (11) where k is the number of iterations, andsis again the shape parameter of the generalized Gaussian dist... |

16 |
Digital image watermarking: hiding information in images
- Kutter
- 1999
(Show Context)
Citation Context ...generalized Gaussian distribution and the watermark as Gaussian. The generalized Gaussian model is given by: p x (x) =s( ) 2( 1s) !N 2 1 j det R x j 1 2 expf( )(jx xj 2 ) T R 2 x jx xj 2 g; (7) 4 (no prior on image) ML-estimate Local Median Laplacian Watermark Local Mean Gaussian Watermark MAP-estimate (with prior on image) Watermark Prediction Wiener (Lee) Filter Gaussian Watermark Gaussia... |

14 |
A copyright protection environment for digital images, Diploma Dissertation, Ecole Polytechnique Federal de
- Perrig
- 1997
(Show Context)
Citation Context ...en equation (6) is reduced to the following minimization problem: ^ x k+1 = arg min ~ x2R N f 1 2 2 n ky ~ x k k 2 + w k+1 kr k k 2 g; (8) where w k+1 = 1 r k 0 (r k ); (9) r k = x k x k k x ; (10) 0 (r) =s[( )]sr jrj 2s; (11) where k is the number of iterations, andsis again the shape parameter of the generalized Gaussian distribution. In this case, the penalty function is quadratic for asx... |

7 |
Podilchuk and Wenjun Zeng. Image-adaptive watermarking using visual models
- Christine
- 1998
(Show Context)
Citation Context ... asxed weighting function w. Assuming w is constant for a particular iteration step, one can write the general RLS solution in the next form: ^ x = w 2 n w 2 n + 2 x x + 2 x w 2 n + 2 x y: (12) 5 This solution is similar to the closed form Wienerslter solution. 9 The same RLS solution could also be rewritten in the form of Leeslter 9 : ^ x = x + 2 x w 2 n + 2 x (y x): (13) The prin... |

4 |
Resolving rightful ownerships with invisible watermarking techniques: limitations, attacks, and implications
- Yeo, Yeung
- 1998
(Show Context)
Citation Context ...oshynovskiy et al. 14 The noise visibility function is dened as: NV F = ! ! + 2 x ; (15) where 2 x the cover image variance, and ! is a local weighting function dened by: ! =s[( )]s1 jrj 2s; (16) where r = x x x . is a tuning parameter dened by: = 100 2 max (17) There NVF can be used for two dierent models. In thesrst model we assume a non-stationary Gaussian cover image. In this c... |

2 |
UnZign watermark removal software. http://altern.org/watermark
- Unknown
- 1997
(Show Context)
Citation Context ... 2 x y: (12) 5 This solution is similar to the closed form Wienerslter solution. 9 The same RLS solution could also be rewritten in the form of Leeslter 9 : ^ x = x + 2 x w 2 n + 2 x (y x): (13) The principal dierence with classical Wiener or Leeslters is the presence of the weighting function w. This weighting function depends on the underlying assumptions about the statistics of the cover... |