University of Ioannina
Νέα στατιστικά μοντέλα εικόνας για την ανίχνευση και σχεδίαση υδατοσημάτων
Abstract
dc:descriptionIn this dissertation we developed new statistical image models for the image watermarking problem. More specifically, we propose stochastic non-stationary image models as image priors necessary for the construction of new watermark detectors for the additive and multiplicative watermarking problems. Also using the image model’s parameters, we design new perceptual masks that help improve watermark detection and robustness, by allowing insertion of imperceptible watermarks with more energy. Additive and multiplicative watermark detection can be formulated as a hypothesis testing problem, where one determines the presence or absence of a known watermark in an image. Within this formulation, the watermark is treated as the known signal and the image is the unknown noise. To derive test statistics for this problem, such as the Likelihood Ratio test detector, a statistical model for the image has to be defined. We propose a hierarchical, two-level image model, which has found applications in image recovery problems. The top level of this model exploits the spatially varying local statistics of the image, while the bottom level is used to characterize the image variations along two principal directions. Based on this model, we derive a class of detectors for the additive watermark detection problem, which include a Generalized Likelihood Ratio, Bayesian, and Rao test detectors. We also propose methods to estimate the necessary parameters for these detectors. Our numerical experiments demonstrate that these new detectors provide superior performance as compared to several state-of-the-art detectors. Next we present new detectors for additive watermarks when the power of the watermark is unknown. These detectors are based on modeling the image using Student’s-t statistics. Using the generative properties of the Student’s-t density function, the Expectation-Maximization algorithm is used to obtain Maximum Likelihood estimates of their parameters. Using these image models detectors based on the Generalized Likelihood Ratio and Rao tests are derived for this problem. Numerical experiments are presented that demonstrate the properties of these detectors and compare them with previously proposed detectors. We also propose a class of watermark detectors based on Total Variation (TV). TV is an important quantity that has found many applications in robust statistics and image recovery. TV is used in this problem through the introduction of a new image prior. This prior uses a spatially weighted version of the TV. The resulting detector is compared to previous similar works and other top-ranking methods for this problem. The proposed detector provides excellent sensitivity in terms of probability of correct detection when the watermark is very weak. Furthermore, it is very robust against most watermark attacks. We also used the proposed hierarchical image prior for the multiplicative watermarking problem where we also propose new watermark detectors. Furthermore, we demonstrate its improved robustness compared to other state of the art similar in spirit watermark detectors. In addition, we propose a new perceptual mask which improves the robustness of additive watermark detectors in the spatial domain. The proposed mask is based on the local image variations along the two principal directions and enhances the watermark’s energy while satisfying the imperceptibility requirement. Numerical results are provided that demonstrate both the value of the proposed mask, and the improved sensitivity as compared to additive watermarking, with the same image model, as the proposed multiplicative watermark detector. Finally, we applied the weighted TV prior to the multiplicative watermarking problem also. We compared the resulting detector with other state of the art multiplicative detectors. Additionally, we propose a novel perceptual mask that exploits the local image characteristics using the variance parameters of the proposed TV prior and emulates the properties of the Human Visual System. Using this mask, we achieved better detection performance as compared to other detectors and improved robustness to intentional or unintentional attacks.
Degree
thesis:*- Grantor dc:publisher
- University of Ioannina
- Year dc:date
- 2010
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Mairgiotis, Antonios
- Μαϊργιώτης, Αντώνιος
Subjects
dc:subject × 22- Υδατοσήμανση
- Υδατογραφία
- Μοντέλα εικόνων
- Ανίχνευση υδατοσημάτων
- Σχεδίαση υδατοσημάτων
- Ένθεση υδατοσημάτων
- Μπεϋζιανές τεχνικές
- Χωρικά προσαρμοζόμενες τεχνικές
- Χωρικά σταθμισμένο Total Variation (συνολικής διακύμανσης) μοντέλο
- Watermarking
- Image models
- Watermark detection
- Watermark design
- Watermark embedding
- Total Variation
- Student-t
- Spatially weighted
- Adaptive methods
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/28309