Loading...

Design and Implementation of the Multiplicative Watermarking Technique for Multimedia Signals

Akhaee, Mohammad Ali | 2009

678 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 40237 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Marvasti, Farrokh
  7. Abstract:
  8. One of the most effective and robust algorithms in watermarking are additive and multiplicative methods. Although the detector of additive watermarking methods are easier than the multiplicative one, they do not gain from human visual or auditory systems. This is the main drawback of additive watermarking techniques. On the other hand, the most advantage of multiplicative watermarking methods is that the power of the watermark is proportional to the power of the host signal. In this thesis, we have introduced a new multiplicative watermarking technique for audio and image signals. For the audio signal, the embedding is performed on the wavelet coefficients. We used Maximum likelihood rule for optimal detection. The multiplicative method is presented in two versions, semi blind and blind. One of the main characteristics of the proposed blind detector is its similar performance for the variety of distributions of the host signals. To control the quality of the watermarked audio, we have used an iterative approach which measures the difference between the clean and the distorted signal. Using this approach, we choose the optimum value for the power of the watermark. The performance of the system is also analytically calculated. In the next step, the multiplicative method is applied to the approximation coefficients of the blocks of the images. By statistical test, we show that these wavelet coefficients can be well modeled by Gaussian distribution. Then we designed the optimal detector and calculated its probability of error precisely. The trade-off between the robustness and transparency of the watermark is elegantly solved via a multi-objective optimization method. To this aim, the image quality index has been used for evaluating the quality of the watermarked image. For the next experiment, the edges of the image are considered for data hiding. Since the wavelet transform cannot sparsely represent the first-order discontinuities of signals, we have employed the contourlet transform for this purpose. Here, to embed a watermark bits, we have used two specific functions rather than a specific value so that more visibility is achieved. By using some approximations, the distribution of watermarked noisy contourlet coefficients are estimated and the optimal detector is implemented. Besides, introduce a synchronization technique to make the algorithm robust to rotation attack. As the last case, we have presented an optimal rational detector for highly correlated signals. It is proved that under some specific circumstances, the ratio variable has far less correlation than the original ones. We used this fact to design a simpler suboptimum detector which causes less computational cost. The proposed method is not vulnerable to gain attack. Experimental results over artificial signals confirm the validity of the resultant relations and confirm the accuracy of our approximations. Extensive simulations over natural signals such as audio and images show that the proposed schemes outperform the recent techniques in the fair and similar conditions.



  9. Keywords:
  10. Wavelet Transform ; Contourlet Transform ; Audio Signals ; Multiplicative Watermarking ; Optimal Detector ; Multimedia Signals

 Digital Object List

 Bookmark

No TOC