【作者】 纪薇；

【导师】 林梓；

【作者基本信息】 吉林大学 ， 信号与信息处理， 2008， 硕士

【摘要】 本文研究了基于混沌加密及HVS(Human Visual System)理论的数字图像盲水印算法,针对水印技术所要兼顾的鲁棒性和不可见性之间的矛盾,在整个嵌入过程中,采用各种技术尽量达到二者之间的折衷。文中进行了大量的仿真实验,实现了二值图像水印和灰度图像水印的嵌入和盲提取,并将各种攻击加入到含水印作品,从遭受攻击后的作品中提取出水印,通过对水印鲁棒性的分析验证算法的可行性。文章首先讨论了二值图像水印的嵌入算法。水印嵌入前,对水印进行混沌加密、置乱处理。嵌入过程利用提升小波变换在频域中进行,采用基于量化的抖动调制算法,并根据HVS理论考虑人眼视觉特性,自适应选取量化步长,完成二值图像水印的嵌入和盲提取。为提高水印的鲁棒性,本文在算法上进行了改进,在嵌入前又对水印进行了纠错编码,提取时进行解码,以便更好的提取出水印。在二值图像水印的基础上,又讨论了灰度图像水印的嵌入,应用位平面分解法嵌入灰度图像水印的重要信息,以兼顾水印的鲁棒性和不可见性。整个算法利用MATLAB仿真环境,实现数字图像水印的嵌入和盲提取,并仿真了水印在经受各种攻击后的提取情况,从实验结果上验证了本文所讨论的数字图像水印算法的可行性。最后,对全文进行了总结和展望。更多还原

【Abstract】 With the growth of multimedia technology and digital network communication,the research of digital information security as well as multimedia copyrightprotection becomes an important issue. As a branch of information hidden, Digitalwatermark technology is an effective mean to resolve the issue of multimedia datacopyright protection.In this paper, digital image blind watermark algorithm based on chaosencryption and HVS theory is in depth study. The watermark embedding andextraction model is established, then the parts of the algorithm of model is intheoretical study. Simulation of the experiment realize the binary image embeddingand extraction. In order to test the feasibility of algorithm, watermark attackedextraction is simulated. Based on the analysis of simulation results, the furtherimproved algorithm is proposed. Compared to previous algorithm, the improvementhas stronger robustness. Finally, these theories is applied to the embedding andextraction of the gray image watermark.According to the traditional watermark technology, scrambling transformationis done before embedding in order to eliminate the spatial correlation among pixels.But the transformation usually has periodicity, the watermark can be restored bycertain iterative times. Based on these considerations, encryption is done beforescrambling transformation in order to enhance the security of watermark. In thispaper, Spatial-temporal chaotic sequence having the characteristic of initial valuesensitivity is used to encrypt. The binary chaotic sequence is generated by OCMLmodel and improved signal function method. Then sequence randomicity isapproved by experiment. The security of watermark can be enhanced by usingchaotic sequence for encryptionThe binary image and gray image are embedded into color image in this paper,the technology of embedding and extraction is adopted as follows:(1) The original color image is transformed into the YIQ color space,watermark is embedded into the Y luminance component of the YIQ.(2) The carrier image is transformed from space field to frequency field bylifting wavelet transform, the watermark is embedded into the wavelet domain.Comparing with the traditional wavelet transform, the lifting wavelet transform has many excellent characteristics, such as simple structure, lower calculation, inversetransform can be inversely realized directly, reversible integer to integer transform,save the memory space and calculation time.(3) The watermark is embedded into the color image using ditheringmodulation algorithm based on quantization. The main idea of algorithm is that thedifferent gray value of carrier image is quantized to the median of correspondingarea according to various quantization algorithms in different quantization condition.The method does not require the original carrier image to extract watermark, it is ablind watermark technique.(4) The HVS theory is applied to this paper in order to select quantization stepadaptively. The carrier image is analyzed according to the characteristic of HVS andmasking effect first, then select small quantization step to the smoothness and lowluminance field, the big quantization step to the texture and high luminance field.Using this theory, the watermark is embedded into the color image throughMATLAB simulative environment. In order to validate the robustness, add manyattacks to watermark image then extract watermark from the attacked image. Theresults show that the algorithm can withstand noise attack, zoom attack and daubingattack especial to the JPEG compress attack, but the ability of resisting circumrotateattack is lower, the robustness of filter attack, cut attack, luminance and contrastchange attack is not higher.Owing to the shortage existing in the algorithm, the improved algorithm isproposed, error-correcting codes theory is added to the former algorithm, this paperadopt hamming code, code before embedding, decode after extracting, the error bitscan be rectified in a certain using this theory, improve the robustness of thealgorithm.The binary image watermark has little information, in order to embed muchinformation, the paper study gray image watermark algorithm based on the binaryimage watermark. Decompose the gray image into many bit planes, the higher bitplanes with important information are embedded into the carrier image. The resultsshow this algorithm has stronger robustness to filter attack, luminance and contrastchange attack, daubing attack especial to noise attack, zoom attack and JPEGcompress attack, but feebler robustness to cut attack and circumrotate attack.To sum up, the paper study the binary image watermark technique, set up theembedding model and extraction model, adopt corresponding technique accordingto the different parts of model. The watermark embedding and extraction is realized through MATLAB simulative environment. The watermark which is attackedextraction is also simulated, According to the results of experiment, the improvedmethod is proposed to improve watermark robustness. At last, the gray imagewatermark algorithm is proposed based on the binary image watermark algorithmand special technology to gray image watermark. The experiment results show thatthe algorithm proposed in this paper has stronger robustness to the noise attack,zoom attack, daubing attack especial to JPEG compress attack, but the algorithm hassome shortage, such as the ability to resist cut attack and circumrotate attack isfeebler, so the author will improve the algorithm in the future.更多还原