I=
Pakistan Research Repository Home
 

Title of Thesis

Digital Watermarking Using Machine Learning Approaches

Author(s)

Imran Usman

Institute/University/Department Details
Department of Computer and Information Sciences / Pakistan Institute of Engineering and Applied Sciences, Islamabad
Session
2010
Subject
Computer Science
Number of Pages
152
Keywords (Extracted from title, table of contents and abstract of thesis)
Digital, Approaches, Machine, Watermarking, Substantial, Illegitimate, Versa, Vice, Perceptual, Optimum, Imperceptibility, Robustness, Tradeoffs

Abstract
In recent years digital watermarking has gained substantial attraction by the research community. It promises the solution to many problems such as content piracy, illicit manipulation of medical/legal documents, content security and so on. Watermarked content is usually vulnerable to a series of attacks in real world scenario. These attacks may be legitimate, such as common signal processing operations, or illegitimate, such as a malicious attempt by an attacker to remove the watermark. A low strength watermark usually possesses high imperceptibility but weak robustness and vice versa. On the other hand, different set of attacks are associated with distinctive watermarking applications,which pose different requirements on a watermarking scheme. Therefore, intelligent approaches are needed to adaptively and judiciously structure the watermark in view of the current application
In addition, traditional watermarking techniques cause irreversible degradation of an image. Although the degradation is perceptually insignificant, it may not be admissible in applications like medical, legal, and military imagery. For applications such as these, it is desirable to extract the embedded information, as well as recover the sensitive host image. This leads us to the use of reversible watermarking. An efficient reversible watermarking scheme should be able to embed more information with less perceptual distortion, and equally, be able to restore the original cover content. Therefore, for reversible watermarking, capacity and imperceptibility are two important properties.However, if one increases the other decreases and vice versa. Hence, one needs to make an optimum choice between these two properties for reversible watermarking.
The research in this work is two-fold. Firstly, we develop intelligent systems for making optimum robustness versus imperceptibility tradeoffs. The performance of the existing watermarking approaches is not up to the task when we consider watermark structuring in view of a sequence of attacks, which is much desirous in real world applications. In order to resist a series of attacks, we employ intelligent selection of both the frequency band as well as strength of alteration for watermark embedding using Genetic Programming. To further enhance the robustness of the watermarking system,Support Vector Machines and Artificial Neural Networks are applied to adaptively modify the decoding strategy in view of the anticipated sequence of attacks at the watermark extraction phase.
Secondly, we devise an intelligent system capable of making optimum/ near optimum tradeoff between watermark payload and imperceptibility. In the context of reversible watermarking, we propose an intelligent scheme which selects suitable coefficients in different wavelet sub-bands and yields superior capacity versus imperceptibility tradeoff. Experimental results show that machine learning approaches are very promising in state of the art watermarking applications.

Download Full Thesis
3,244 KB
S. No. Chapter Title of the Chapters Page Size (KB)
1 0 CONTENTS
 

 

14
41 KB
2 1 INTRODUCTION

1.1 Motivation
1.2 Watermarking in Historical Perspective
1.3 Robust Image Watermarking
1.4 Reversible Watermarking
1.5 Research Objectives and Contributions
1.6 Structure of the Thesis

23
61 KB
3 2 DIGITAL WATERMARKING AND MACHINE LEARNING: SOME PRELIMINARIES

2.1 Digital Watermarking
2.2 Human Visual System Modeling
2.3 Additive Spread Spectrum Watermarking (ASSW)
2.4 Error Correcting Codes and Digital Watermarking
2.5 Attacks and their Countermeasures
2.6 Genetic Programming and the Basic GP Algorithm

35
167 KB
4 3 ATTACK RESISTANT VISUAL TUNING

3.1 Attack Resistant Visual Tuning: The Basic Architecture
3.2 Watermark Embedding
3.3 Generating Candidate VTF in GP Domain
3.4 Watermark Extraction
3.5 Evaluating Each Potential VTF
3.6 Implementation details
3.7 Potential applications of the IARW approach
3.8 Experimental results and discussion
3.9 Summary

59
1,727 KB
5 4 INTELLIGENT EMBEDDING AND EXTRACTION: A NEW PARADIGM IN WATERMARK ENCODING AND DECODING STRUCTURES

4.1 Introduction
4.2 GP based Watermark Embedding Phase
4.3 Computational Intelligence based Adaptive Watermark Decoding
4.4 Security Enhancement
4.5 Potential Applications
4.6 Results and Discussion
4.7 Summary

92
545 KB
6 5 REVERSIBLE WATERMARKING USING MACHINE LEARNING

5.1 Introduction
5.2 Intelligent Reversible Watermarking
5.3 Results and Discussion
5.4 Prospective Applications of the Proposed Scheme
5.5 Summary

125
494 KB
7 6 CONCLUSIONS AND FUTURE DIRECTIONS

6.1 Intelligent Watermark Tuning
6.2 Intelligent Embedding and Extraction
6.3 Reversible Watermarking using Machine Learning
6.4 Future Directions

140
42 KB
8 7 REFERENCES

 

148
36 KB