Optimisation of Tamper Localisation and Recovery Watermarking Techniques

Digital watermarking has found many applications in many fields, such as:
copyright tracking, media authentication, tamper localisation and recovery,
hardware control, and data hiding. The idea of digital watermarking is to embed
arbitrary data inside a multimedia cover without affecting the perceptibility of the
multimedia cover itself. The main advantage of using digital watermarking over
other techniques, such as signature based techniques, is that the watermark is
embedded into the multimedia cover itself and will not be removed even with the
format change.
Image watermarking techniques are categorised according to their robustness
against modification into: fragile, semi-fragile, and robust watermarking. In fragile
watermarking any change to the image will affect the watermark, this makes fragile
watermarking very useful in image authentication applications, as in medical and
forensic fields, where any tampering of the image is: detected, localised, and
possibly recovered. Fragile watermarking techniques are also characterised by a
higher capacity when compared to semi-fragile and robust watermarking. Semifragile
watermarking techniques resist some modifications, such as lossy
compression and low pass filtering. Semi-fragile watermarking can be used in
authentication and copyright validation applications whenever the amount of
embedded information is small and the expected modifications are not severe.
Robust watermarking techniques are supposed to withstand more severe
modifications, such as rotation and geometrical bending. Robust watermarking is
used in copyright validation applications, where copyright information in the image
must remains accessible even after severe modification.
This research focuses on the application of image watermarking in tamper
localisation and recovery and it aims to provide optimisation for some of its
aspects. The optimisation aims to produce watermarking techniques that enhance
one or more of the following aspects: consuming less payload, having better
recovery quality, recovering larger tampered area, requiring less calculations, and
being robust against the different counterfeiting attacks. Through the survey of the main existing techniques, it was found that most of them
are using two separate sets of data for the localisation and the recovery of the
tampered area, which is considered as a redundancy. The main focus in this
research is to investigate employing image filtering techniques in order to use only
one set of data for both purposes, leading to a reduced redundancy in the
watermark embedding and enhanced capacity. Four tamper localisation and
recovery techniques were proposed, three of them use one set of data for
localisation and recovery while the fourth one is designed to be optimised and
gives a better performance even though it uses separate sets of data for
localisation and recovery.
The four techniques were analysed and compared to two recent techniques in the
literature. The performance of the proposed techniques vary from one technique to
another. The fourth technique shows the best results regarding recovery quality
and Probability of False Acceptance (PFA) when compared to the other proposed
techniques and the two techniques in the literature, also, all proposed techniques
show better recovery quality when compared to the two techniques in the
literature.