الفهرس 
coverOnly 14 pages are availabe for public view
 |  | from | 160 |  |  |
| | | from | 160 | | |
Abstract
The presented work focuses on using digital watermarking for image verification and authentication. This entails elaborate presentation, implementation, performance evaluation, and comparison of known watermarking techniques. Suggestion of possible improvement, enhancements, or upgrades of studied techniques falls within the scope of interest of the thesis. All investigations are based on extensive Monte-Carlo simulations.
We study closely four typical single-bit watermarking systems that differ in the embedding algorithm, the detection algorithm, or both. Through elaborate presentation, implementation, and performance evaluation, we show the positive and the negative merits of each system. We address important issues such as selection of an embedding strength satisfying the fidelity requirement, evaluation of potential detection performance and analysis of trade-offs among fidelity, robustness and detection performance of each considered system.
The thesis also investigates multibit-watermarking systems and considers methods of representing messages with watermarks. We elaborate on two approaches to modulating messages. These are direct message modulation, and multiplexing. We study an eight-bit code division multiplexing watermarking system with blind embedding and linear correlation detection. We show that there is a trade-off between achieving good performance and sustaining an acceptable distortion. We show also how error correction codes can much reduce the error probability of the system. Moreover, we investigate the effect of informed embedding on system performance. We also investigate a direct message modulation DMM watermarking system with blind embedding and linear correlation detection. We suggest the use of whitening to improve detection performance if the spatial covariance of images is known. We also develop approaches to implement and to examine the effect of informed embedding and informed modulation on the performance of the system. Results show that informed embedding and modulation improves detection performance over informed embedding but is outperformed by whitening.
We also focus on the use of digital watermarking for exact content authentication. A blind embedder that uses ordinary addition, followed by rounding and clipping does not allow exact recovery of the original image. One solution is to let the embedder apply only rounding and modulo addition. However, this solution introduces a ‘salt-and-pepper’ noise that degrades the detection performance. We suggest a gray-level unwrapping algorithm to alleviate the bad impact of the modulo addition on the detection of the modulo-added erasable watermarks. Computer simulations demonstrate the efficacy of the suggested unwrapping algorithm.
We conclude the thesis with a discussion of watermark immunity, supported by an investigation of three algorithms for watermark removal, two published and one contributed by the author, to estimate a watermark, created using the single-bit blind embedder. Evaluation of the relative perceptual distortion of the image obtained after removal of the estimated watermark confirmed good estimation quality and a successful removal of the mark.