The Research On Video Data Hiding Algorithms Based On H.264/AVC

The security requirement of secret information transmission is becoming stricter in digital age. The traditional cryptography technology encrypts the protected information to get the ciphertext which is hard for understanding. Though the ciphertext is hard to understand, it would leak the existence of encryption behavior and absorb the attackers' attention. The attackers would intercept, break or replay the ciphertext and bring a serious threat to the security of the secret information transmission.Data Hiding is trying to change this situation. It can make the existence of secret information unnoticeable by hiding it into those common files. Using public (non-dedicated facilities) digital covers to transmit information secretly can protect not only the content of the information but also the secret behaviors (encryption and transmission). In recent years, H.264/AVC video encode/decode standard has gained popularity rapidly for its high compression efficiency and outstanding network-friendly property, and has become one of the most conventional ways for video transmission. Thus, it is a brand-new research on secure and robust data hiding method by utilizing widely-used H.264/AVC videos to protect the secret information transmission security in unsafe circumstance.The intra-frame distortion drift problem has been raised to be the key issue of the covert communication utilizing H.264/AVC video covers. Based on the summary of the existing H.264/AVC video data hiding methods, the readable and built-in type methods which hide data into the quantized DCT (Discrete Cosine Transformation) coefficients are considered appropriate for covert communication. The embedding error formula of the luminance block residual deduced from the theoretical analysis of the encoding/decoding and data hiding procedures shows that the embedding error within a block is only related to the location of the embedding coefficients and basically independent on the specific video cover. The embedding error within a block would propogate to other neighbored blocks and accumulates based on the analysis of the intra-frame prediction. This phenomenon is called the intra-frame distortion drift. Analysis of the accumulating procedure and the experimental results shows that the drift problem would bring significant distortion to the embedded videos. This would challenge the visual concealment of the embedded video and exposes the hiding behavior of the secret information consequently. Furthermore, this would limit the embedding capacity and makes the security and robustness enhancement methods inapplicable. Thus, in order to utilizing H.264/AVC video covers for covert communication, it is crucial to solve the intra-frame drift problem.The coupled relationship of the coefficients'locations in IntDCT (Integer Discrete Cosine Transform) is discovered.In order to constraint the influence of the intra-frame distortion drift, aiming to eliminate or reduce the embedding error, under the zero assumption of the crucial pixels (which will be used in the intra-frame prediction of other blocks and make the embedding error propagate to other blocks, the pixels on the last line and the last column, totaled 7 pixels), the coupled relationship of the coefficients' locations in IntDCT is discovered. Consequently, the coupled paired-coefficients in IntDCT is defined and 24 coupled paired-coefficients are derived.The coupled-paired coefficients data hiding method is proposed based on the embedding locations. The information is embedded into the embedding coefficient of a coupled-paired coefficients, and the compensation coefficient of the coupled-paired coefficients is modified simultaneously to compensate the embedding error within the block. Specifically, the embedding error of some crucial pixels is compensated to be zero; the propagated error and the accumulated error are reduced consequently. The experimental results show that this method can control and reduce the intra-frame distortion drift.Two kinds of restrictions on the intra-frame prediction modes of the neighbored blocks are proposed to intercept the embedding error propagation. Further analysis on the intra-frame distortion drift shows that the embedding error of current block would propagate to its four neighbored blocks, and the intra-frame prediction modes of these blocks determine whether the embedding error will propagate. Based on the nine intra-frame prediction modes, the vertical-like texture restriction and horizontal-like texture restriction are proposed to intercept the embedding error propagation of crucial pixels on the last row and on the last column respectively. And this can totally avoid the intra-frame distortion drift.The H.264/AVC video data hiding method based on the texture property is proposed. The embedded block and paired-coefficients are chose according to the intra-frame prediction modes of neighbored blocks. Combining the paired-coefficients with the texture property can make the embedding error of some crucial pixels to be zeroes and exclude the other crucial pixels with non-zero embedding error from the referenced pixels of the intra-frame prediction in other blocks. The research proves that this method can deprive the propagation chance of embedding error, break off the accumulation channel and eliminate intra-frame distortion drift completely. The experimental results and comparison with other methods demonstrate the superiority of this method. Its outstanding performance on the embedding capacity and visual concealment makes the security and robustness enhancement methods applicable. Further improvements include stair-like embedding; single-coefficient embedding et al. Besides, the layered robustness enhancement concept is discussed and the embedded bits survive ratio is defined to evaluate the robustness.The single-coefficient embedding method is applied to the reversible data hiding of H.264/AVC videos. Generally, the data hiding covers will go through extensive and severe modifications. However, this is not appropriate for H.264/AVC videos because of the intra-frame distortion drift. The single-coefficient embedding method which can avoid the drift is combined with the reversible data hiding based on difference expansion. The proposed method embeds data into the single coefficient and recovers the original videos at the extraction stage. The experimental results and comparisons to the existing methods prove its correctness and high efficiency. This attempt can also demonstrate the important significance of the breakthrough in the intra-frame distortion drift.