Optimal Coding Theory And Method Study For Reversible Data Hiding

In the last two decades, due to the advances in generation, storage, and commu-nication technology of digital content, and the spring up of the Internet techniques, digital images, electronic documents, digital audio and video medias are already taking the place of the traditional media through which people share and deliver information nowadays. These digital medias are widely stored in kinds of electronic equipments and transmitted through the Internet, which facilitates people’s life vastly. However, the digitization of traditional media and the convenient network transmission bring the data security and integrity a severe test, such as the proliferation of the pirated media and the illegal information theft and tampering, etc. Reversible Data Hiding technique is an promising solution for the data security problem. Through a cover method, reversible data hiding embeds some useful information (like copyright information) into the public released media, so as to achieve the purpose of data forensics and authentication.Reversible Data Hiding is a special data hiding technique with the characteristic that not only the message needs to be precisely extracted, but also the cover itself should be restored losslessly. This property is important in some special scenarios such as medial imagery annotation, military imagery and law forensics. In these applications, the cover is too precious or important to be damaged, otherwise the medical diagnosis or the action at law will fail. Secondly, reversible data hiding can totally restore the cover image after the identification of the image content’s facticity and integrity, which attracts more and more attention these days.Based on the information theory model of the reversible data hiding, the disser-tation discussed the rate distortion problem of reversible data hiding under kinds of application scenarios.The main research work and contributions of the dissertation can be summarized as the following aspects:1. Proposed and constructed a fast estimation algorithm to solve the square error distortion based information theory model of reversible data hiding, deduced the corresponding Lagrange dual problem of the modelBased on the information theory model of reversible data hiding under the square error distortion scenario, the dissertation deduced the corresponding Lagrange dual problem, and then proposed a fast algorithm to estimate the optimal solu-tion. The fast algorithm proposed by the dissertation to solve the rate distortion problem for reversible data hiding not only runs five to10times faster than the existing algorithm, but also is more applicable for practical distortion limited or embedding capacity limited scenarios, which leads to a better solution for realistic reversible data hiding problems.2. Proposed a recursive code construction based on the lossless entropy coding algorithms, realized the optimal histogram modification to approach the up-per bound of the rate distortion model for reversible data hiding, under the square error distortion scenarioBased on the rate distortion model of reversible data hiding under the square error distortion measurement, the dissertation proposed a recursive code construction through utilizing lossless arithmetic coding after estimation the optimal solution of the model. The proposed recursive code construction has been proved that it can approach the rate distortion upper bound infinitely, as long as the cover sequence is long enough. Taking the general nature images as cover sequence, the recursive code construction earns a1.0dB to2.0dB performance gain compared with classical state-of-arts embedding methods.3. Proposed the extension algorithms for the fast estimation algorithm and the recursive code construction to the general distortion measurement cases to make them more suitable for general reversible data hiding applicationsBased on the rate distortion model of reversible data hiding, the dissertation fur-ther extended the fast estimation algorithm and the recursive code construction to the general distortion measurement cases to make them more suitable for general reversible data hiding problems. After the extension, the dissertation discusses kinds of application scenarios, such as binary image reversible data hiding, two dimensional or even four dimensional reversible data hiding cases, which makes the rate distortion model for reversible data hiding more applicable and scalable.4. Proposed a novel optimal evaluation criteria for the prediction part of re-versible data hiding, from the point of view of the rate distortion model, which to a certain extent establish the consistence between reversible data hiding and lossless data compressionBased on the recursive code construction for the rate distortion model of reversible data hiding, the dissertation proposed a novel optimal evaluation criteria for the prediction part of reversible data hiding, which to a certain extent establish the consistence between reversible data hiding and lossless data compression. The dissertation proposed a minimum rate prediction scheme for reversible data hid-ing, which seamlessly connect the front prediction part and the back end embed-ding part in essence to achieve a better global embedding performance. Combined with the optimal recursive code construction, the minimum rate prediction method earns a1.5dB to2.0dB performance gain compared with classical reversible data hiding prediction methods for general nature test images.The last of the dissertation summarizes the whole research work and point out future directions for the continuing work.