|M.Sc Student||Maor Alina|
|Subject||On joint Information Embedding and Data Compression|
|Department||Department of Electrical Engineering||Supervisor||Professor Neri Merhav|
The subject of watermarking and information embedding has been attracting a vast amount of attention of both the academic world and the industry, due to an increasing awareness for the need of data protection in its various forms: ownership identification, data forgery exposure, etc. In this thesis, we examine joint watermarking and lossy compression of the watermarked data (stegotext), where compression of the stegotext is treated as an integral part of the watermarking scheme. The general problem is as follows: There is a set of watermarking messages to be embedded in the source data (covertext) subject to some distortion constraint. The stegotext resulting from this embedding is compressed losslessly and the watermark must be reliably decodable with or without access to the source data, either directly from the stegotext or from its forgery, produced by a stationary memoryless attack channel operating on the stegotext. We characterize the best achievable tradeoffs between the embedding rate, the composite rate and the allowable average distortion, for public watermarking, where the covertext is available to the encoder only. The attack-free case of the problem is treated separately from the general case of an attack operating on the stegatext, due to its conceptual significance and its special qualities. Our analysis is performed for a case of finite-alphabets, and we further argue that the achievable rate region of the continuous case is given by the same expression as in the finite-alphabet case. For the attack-free case of private watermarking, we introduce a coding scheme of higher embedding capacity, which exploits the knowledge of the covertext to both the encoder and decoder during the compression/decompression stage, and derive a single-letter expression of this embedding capacity.