The Research On Reversible Data Hiding Techniques For3D H.264Video

The world was shocked by Prism scandal, which showed that the security problems of network information need to be solved. Encrypted information is hard to be interpreted, but the exposed encryption behavior will cause the third party’s attention easily. Both the information and the pass behavior are hidden by using data hiding technology to embed information into the mass media. Non-reversible information hiding techniques will cause permanent distortion of the carriers, but when reversible data hiding techniques are used to hide information, the original carriers can be completely recovered after extracting information. In this way, the useful multimedia could be saved and the risk of being cracked in the future can be eliminated. When information is hidden into3D H.264videos reversibly, there are three problems. First, prediction techniques are used for coding and decoding3D H.264videos, just a modified block may cause the change of other views, frames and blocks in the same frame, which are called inter-view distortion drift, inter-frame distortion drift, and intra-frame distortion drift. Second, in the traditional histogram shifting algorithm, the value of each coefficient is individually modified to embed one bit of information so that video quality, embedding capacity and efficiency are limited. Thirdly, the random selection space of the general reversible information hiding algorithms is small, so they are easy to crack. Therefore, in order to solve these problems, reversible information hiding technology for3D H.264video is researched.The methods for limiting distortion drift are proposed. Inter-view prediction, inter-frame prediction, and intra-frame prediction are applied in3D H.264video coder and decoder, thus if only a quantized discrete cosine transform (QDCT) luminance coefficient is simply modified for embedding information, it may cause inter-view distortion drift, inter-frame distortion drift and intra-frame distortion drift. Therefore, according to the prediction structure of3D H.264video, the ways of preventing the inter-view and the inter-frame distortion drift is presented. There is no inter-view distortion drift when data is embedded into the frames that do not predict other views. Inter-frame distortion drift will not be caused by hiding data into a frame that is not referred by other frames in the same view. Furthermore, according to the principle of intra-frame distortion drift, three ways of avoiding intra-frame distortion drift are proposed.(1) The distortion drift is limited in the frames that have inter-frame or inter-view predicted macro blocks. In the4×4transform macro block with inter-frame or inter-view prediction, if4×4blocks that are not at the bottom row or the rightmost column are selected for embedding information, intra-frame distortion drift will not be caused.(2) Intra-frame distortion drift is limited according to the macro block type of neighboring blocks. If intra-frame prediction is not used by the right and the under adjacent blocks of the current block, the current block could be chosen as an embeddable block.(3) Intra-frame distortion drift is prevented in any frame. Embeddable blocks are selected according to the macro block type or the intra-frame prediction modes of the neighboring blocks. Experimental results show that these methods are effective in limiting distortion drift. Compared with other methods, higher video quality, embedding capacity and efficiency could be obtained by using the three methods of limiting intra-frame distortion drift.Two reversible data hiding algorithms based on2D histogram shifting by associating dual QDCT luminance coefficients are proposed. One is a reversible algorithm by which multiple bits of information are hidden in the associated dual coefficients, and the other one is a reversible algorithm by which two bits of data can be embedded in the combined dual coefficients. In traditional histogram shifting algorithm, one bit of information is embedded by modifying a single coefficient value without considering the associated coefficients. In order to further improve the embedding capacity and reduce distortion, novel2D two histogram shifting algorithms uniting dual coefficients are proposed. In one embeddable block limiting distortion drift, two QDCT alternating current (AC) coefficients are randomly selected as a pair of embeddable coefficients. A coefficient histogram is constituted by all embeddable coefficient pairs selected from the video. The values of coefficient pairs are divided into disjoint sets. According to the set which the value of coefficient pair belongs to, data could be hidden by expanding or shifting the2D coefficient histogram. By jointing the two luminance coefficients for embedding information, in most cases, when the luminance value of one QDCT coefficient is changed by adding or subtracting1at most, three bits of information could be simultaneously embedded by using the first algorithm, and two bits of information could be simultaneously hidden by using the second algorithm. Experimental results show that compared with traditional2D histogram shifting algorithm, higher embedding capacity, video quality and embedding efficiency can be achieved by using the first algorithm; in order to embed the same amount of information, higher embedding efficiency and less distortion can also be obtained by using the second algorithm.Two reversible information hiding algorithms based on3D histogram shifting by uniting multiple coefficients are proposed. One is a reversible algorithm for hiding fixed-length information, and the other one is a reversible algorithm for hiding variable length information. In order to break the limitation caused by the single modification of traditional histogram shifting algorithm,3D histogram shifting algorithms by associating multiple coefficients are proposed. In a4×4luminance embeddable block limiting distortion drift, three QDCT AC coefficients are selected as an embeddable unit. The values of coefficient units are divided into disjoint sets. According to the set which the values of coefficient units belong to, the information is hidden and the carrier is modified. In order to embed two bits of information, in most cases, only the value of one QDCT coefficient is modified by adding1at most. When information is embedded by using the first algorithm, the same bits of information will be hidden in coefficient units that belong to the same set. The second algorithm is based not only on the value of the coefficient unit, but also the embedded information. When the value of the coefficient unit is fixed, the length of the embedded information will vary with different values of information. Experimental results show that compared with some other reversible information hiding algorithms, when the same amount of information is embedded, better video quality and embedding efficiency can be got by using the two proposed reversible information hiding algorithms.A reversible information hiding algorithm based on orthogonal vectors is proposed. In general reversible information hiding algorithm, the random selection space is so small that the information will be quickly extracted if a third party gets carrier with information. In order to enhance the ability of anti-steganalysis, a reversible information hiding algorithm with a large random space is proposed. A non-repeating random sequence is generated by using a random seed. A carrier vector is composed of the QDCT AC luminance coefficients that are correspondent with the random sequence. An offset vector with a smaller modulus value is generated randomly (the offset vector and the carrier vector have the same number of elements). The value of the inner product of the carrier vector and the offset vector is divided into several disjoint intervals. Data is hidden based on the interval which the inner product value belongs to. When a to-be-hidden bit is0, the carrier vector is not changed. Otherwise, according to the sign of the inner product of the two vectors, the offset vector is adding to or subtracting from the carrier vector. The receiver could construct the carrier vector and the offset vector by using the same random seed, and extract information and recover the carrier based on the interval of inner product of the two vectors. Over1.0×1014ways could be used to constitute the carrier vector and the offset vector, which increases the difficulty of stealing information. Experimental results show that compared with several reversible data hiding algorithms, better video quality and undetectability can be achieved by using this proposed algorithm.