Research On Information Hiding Algorithm Based On Human Vision

With the rapid development of the information technology and computer network, information transmission and processing are becoming more and more convenient. But information security has come to be a social issue at the same time. At present, information security includes two techniques:Cryptography and Information Hiding. As to cryptography, the secret information is transformed into a random sequence by encrypting, only the partners with the key are able to correctly extract the information. But with the development of cryptanalysis and the high speed of computer parallel processing technology, the traditional password security system is faced with severe challenges, and garbled encrypted information is more easily leading to the attention of the preventer. Even if the preventer can not break in, the successful intercepted information may also be damaged and interfered with normal communications.Limitations of the cryptography promote the generation and development of the information hiding technology. As an important method of information security transmission, information hiding technology hasattracted great concern. By this means, the specific information is embedded into digital carrier information, and then the confidential information is transferred through the public one. It aims not to restrict normal access to information, but rather to ensure that hidden information does not draw monitor’s attention, so the likelihood of attack is reduced. Compared with cryptography, information hiding not only hides the content of information but also hides the existence of information. The security comes from t the paralysis perception of the third parties, making secure communication from the "readable" to "invisible ". Therefore, in some cases, information hiding is more secure than encryption.For information hiding techniques, there are three most concerned points: imperceptibility, hidden capacity and robustness. The three evaluation aspects are mutually interdependent, while not optimal. So as to a specific application an appropriate balance should be sought. For the general information hiding, the imperceptibility and hidden capacity are more important. For the secret information hiding, hthe imperceptibility and robustness are more important. In the thesis three aspects are discussed based on the image and video:capacity improvement, reducing image distortion and robustness improvement. The embedding domain feature extraction and adaptive embedding algorithm based on human visual characteristics are studied. Several information hiding algorithms are proposed based on different carriers and feature extraction.The main contributions are as following:(1) A fresh information hiding algorithm based on Stentiford visual attention model is proposed. Traditional information hiding algorithms improved the invisibility by the means of image vision masking, rarelyfby the selection of the attention region. This thesis, Stentiford visual attention model is utilized to diverse the region according to the degree of attention. By embedding less information in sensitive regions, the invisibility of a image with hidden information is improved. By embedding more information in non-sensitive regions, the information capacity is improved. Experiments demonstrate that the proposed algorithm not only achieves good performance on information capacity and visual invisibility, but also improves the robustness. Meanwhile, a visual attention based PSNR algorithm is also proposed. It provides a reference to the subjective visual evaluation of image quality.(2) An Adaptive Quantization Index Modulation (AQIM) algorithm based on the Watson’s visual model is proposed. Secret information is invisible during the transmission process. Hidden information should be extracted even if the carrier image is processed and attacked. Then the authentication and integrity can be enhanced. In the thesis, information hiding algorithm is studied based on the visual hidden attribute. The defaults of the Adaptive QIM algorithm are analyzed first. Then an improved iterative AQIM algorithm is proposed.By combining the two factors of information hiding embedding domain selection and embedding algorithm, we proposed an information hiding scheme which combines the Stentiford visual attention model and iterative AQIM algorithm. In this algorithm, the hiding performance is improved from the point view of the attention region selection and region detail. Experiments demonstrate that the proposed algorithm has better imperceptibility and robustness.ICA is a kind of signal processing method that can extract the independent characteristics of the signal and the characteristics are independent in the sense of the statistics. Because the maximum embedding capacity can be obtained in the independent information hiding embedding domains, an information hiding scheme is proposed by combining subsample ICA and iterative AQIM. Experiments demonstrate that the proposed algorithm has better statistical property and robustness.(3) A video information hiding scheme is proposed which has high capacity and robustness to MPEG2, MPEG4 to H.264 transcoding. With the development of Triple Play Service, decompression and recompression are not important, while the transformation among different compression standards should be more considered. The coding processing of the most standards is based on the block DCT transform. In the proposed algorithm, the spatial relationship between different-size blocks is analyzed first. The DCT coefficients which have less impact to compression and transcoding are selected for information hiding. Compared with the previous algorithm, the proposed one improves the capacity much and has good robustness and invisibility.