半色调图像及视频信息隐藏算法研究
发布时间：2018-07-17 08:09
【摘要】：随着计算机和网络技术的不断发展,人们在享受网上大量资源的同时,如何保证信息的安全已经成为一个亟需解决的问题。信息隐藏是信息安全领域的一项主要技术。图像和视频信息隐藏成为最常用的形式,一般指通过特定方式对图像和视频的数据进行修改,从而隐藏信息,在隐藏后,图像和视频的使用不受影响。半色调图像是一种特殊的图像载体,仅通过黑白两个颜色值来表示一幅图像。半色调图像中信息隐藏可分为两类:(1)网点水印:一幅半色调图像中,通过网点特征,直接隐藏水印信息;(2)半色调可视加密(halftone visual cryptography,简称HVC):在多幅半色调图像中,隐藏一幅秘密图案。在HVC中,(2,2) HVC最常用,即将一幅秘密图案隐藏在两幅半色调图像中,通过处理两幅半色调图像,可提取出秘密图案。目前已有的(2,2) HVC算法,在提取图案的正确率(CDR)和生成图像的质量上,都有待提高。另外,现有(k,n) HVC的算法,需要码书,并会造成像素扩展,即随着用户数目n的增加,生成半色调图像的分辨率需要扩展为原图像的很多倍。立体视频是一种特殊的视频载体,每帧立体视频包括左右两个通道的图像,这两个通道的图像相近,但不相同,立体视频水印成为一个新的视频水印研究方向。由于实际项目需要,希望实现一种能够抵抗多种攻击的图像水印算法,包括:编码攻击、裁剪攻击、直线攻击、缩放攻击,并且能保证一定的水印容量。针对以上的各种问题,本文针对半色调信息隐藏、立体视频水印和图像水印进行研究,主要取得以下成果:(1)基于新共轭的彩色(2,2)半色调可视加密算法。针对彩色半色调图像,提出新共轭的概念,并将这个概念应用到误差扩散半色调图像和点扩散半色调图像信息隐藏中,提出基于误差扩散的新共轭数据隐藏算法(Data hiding by new conjugate color error diffusion,简称NCCED)和基于点扩散的新共轭数据隐藏算法(Data hiding by new conjugate color dot diffusion,简称NCCDD )。两种算法将秘密图案隐藏在两幅彩色半色调图像中。通过执行这两种算法,在彩色半色调图像信息隐藏过程中,减少需要置换的分量数目,减少由于置换干扰太大而无法实施数据隐藏的情况,可以提高提取秘密图案的正确解码率(CDR),还可提高生成半色调图像的质量;提出平均每像素干扰值(Average distortion per pixel,简称 ADPP)的概念,通过 ADPP 的大小衡量信息隐藏算法对生成半色调图像质量的影响。(2)基于块共轭的高质量(2,2)半色调可视加密算法。选择多个半色调图像作为训练样本,生成最小均方滤波器(Least mean square filter,简称LMSF),用LMSF评价半色调图像的质量。为提高信息隐藏算法生成半色调图像的质量,提出块共轭的概念,将块共轭概念应用到半色调信息隐藏中,提出一种基于块共轭的高质量半色调信息隐藏算法(High quality data hiding in halftone images based on block conjugate,简称HQDHBC)。该算法将一幅秘密图案隐藏在两幅基于误差扩散的半色调图像中。在两幅图像对应的M×N分块中,如果前M×N-1个点都共轭或相同时,则最后一个点不影响块共轭关系,最后一点执行正常的半色调方法,而不执行信息隐藏算法。由于执行HQDHBC算法,可在不影响秘密图案提取正确率的前提下,减少实施信息隐藏算法的点数,从而提高生成半色调图像的质量。(3)基于Shamir分享的(k,n)半色调可视加密算法。将Shamir秘密分享应用到半色调可视加密中,提出一种基于Shamir秘密分享的(k,n)半色调可视加密算法(a (k,n) halftone visual cryptography based on Shamir's secret sharing,简称 HVCSSS)。通过本算法,可实现(k,n)半色调可视加密,即将一幅二值秘密图像隐藏到n幅半色调图像中,任意选取k幅半色调图像,都可以提取原始的秘密图像。首先,利用Shamir分享,将原始二值秘密图像,分享为n幅无意义分享图像;然后,利用半色调图像的自隐藏技术,将分享图像隐藏在半色调图像中,构成n幅半色调分享图像;第三,从n幅半色调分享图像中,通过半色调图像自提取技术,提取出n幅隐藏的分享图像;最后,从n幅提取出的分享图像中,任意选择k幅分享图像,利用拉格朗日插值技术,提取隐藏的秘密图像。(4)变换域立体视频盲水印算法。在实现并分析基于DCT和基于DFT的视频水印算法性能基础上,提出一种变换域立体视频盲水印算法。通过立体视频左右路图像中分别嵌入具有相反嵌入规则的DCT水印和DFT水印,在视频受到攻击后,结合左右路图像相对关系提取水印,并采用择优使用的方式从DCT水印和DFT水印中选择提取鲁棒性更强的水印。(5)基于DFT的高效改进图像水印算法。在实现并分析基于DFT的图像水印算法基础上,提出一种基于DFT的高效图像水印算法。对原始水印信息进行混合纠错编码,保证嵌入水印后的图像,在承受攻击情况下,提取正确率达到100%;针对每个4×4频域数据块改变嵌入水印方式,分别采用右上-左下和左上-右下的能量关系表示水印值,并且用2×1的频域点来表示能量关系,可在PSNR基本相同的前提下,嵌入容量增加为原来的2倍;通过调整嵌入水印的范围,扩展水印容量。
[Abstract]:With the continuous development of computer and network technology, while people enjoy a lot of resources on the Internet, how to ensure the security of information has become an urgent problem. Information hiding is a major technology in the field of information security. Image and video information hiding is the most commonly used form, generally refers to a specific way to the image. The data of video and video are modified to hide information. After hiding, the use of images and videos is not affected. The halftone image is a special image carrier, only by two color values of black and white to represent an image. The information hiding in the halftone image can be divided into two categories: (1) a dot watermark: a halftone image through a dot. Features, direct hidden watermark information; (2) halftone visual encryption (halftone visual cryptography, HVC): in multiple halftone images, hidden a secret pattern. In HVC, (2,2) HVC is most commonly used, a secret pattern is hidden in two halftone images, and the secret pattern can be extracted by processing two halftone images. The former (2,2) HVC algorithm needs to be improved in extracting the correctness of the pattern (CDR) and generating the quality of the image. In addition, the existing (k, n) HVC algorithm needs codebook and will cause pixel expansion, that is, with the increase of the number of users n, the resolution of the generated half tone image needs to be extended to many times the original image. The stereo video is a kind of one. Special video carrier, each frame video includes images of about two channels. The images of the two channels are similar, but different. Stereo video watermarking becomes a new research direction of video watermarking. Because of the actual needs of the project, we hope to achieve a kind of image watermarking algorithm that can resist many attacks, including: coding attack, cutting attack. Attack, straight line attack, zoom attack, and can guarantee a certain watermark capacity. Aiming at all the problems above, this paper focuses on halftone information hiding, stereo video watermarking and image watermarking. The main achievements are as follows: (1) a new conjugate based color (2,2) halftone visual encryption algorithm. A new method for color halftone images is proposed. A new conjugate data hiding algorithm based on error diffusion (Data hiding by new conjugate color error diffusion, abbreviated NCCED) and a new conjugate data hiding algorithm based on point diffusion (Data hiding by) are applied to the concept of conjugation, which is applied to error diffusion half tone image and point diffusion half tone image information hiding. Conjugate color dot diffusion, abbreviated as NCCDD). The two algorithms hide the secret pattern in two color half tone images. By executing these two algorithms, the number of components needed to be replaced in the process of color half tone image information hiding, and the situation that the data hiding can not be implemented because the displacement interference is too large and can be improved. The correct decoding rate (CDR) of the secret pattern can also be extracted, and the quality of the generated half tone image can be improved; the concept of average per pixel interference value (Average distortion per pixel, abbreviated for short) is proposed, and the influence of the information hiding algorithm on the quality of the generated halftone image is measured by the size of ADPP. (2) high quality (2,2) halftone based on block conjugation According to the encryption algorithm, multiple half tone images are selected as training samples, and the minimum mean square filter (Least mean square filter, called LMSF) is generated to evaluate the quality of the halftone image with LMSF. In order to improve the quality of the half tone image generated by the information hiding algorithm, the concept of block conjugation is proposed and the block conjugation concept is applied to the halftone information hiding. A high quality halftone information hiding algorithm based on block conjugation (High quality data hiding in halftone images based on block conjugate, referred to as HQDHBC) is proposed. The algorithm hides a secret pattern in two halftone images based on error diffusion. When the conjugate or the same, the last point does not affect the block conjugate relation, the last point executes the normal halftone method and does not carry out the information hiding algorithm. Because of the execution of the HQDHBC algorithm, the number of information hiding algorithms can be reduced without affecting the accuracy of the secret pattern extraction, thus improving the quality of the generated halftone image. (3) the (k, n) half tone visual encryption algorithm based on Shamir sharing. The Shamir secret sharing is applied to the halftone visual encryption, and a K, n semi tone visual encryption algorithm based on Shamir secret sharing is proposed. Halftone visual encryption, a two value secret image is hidden in the N amplitude and half tone image, and K amplitude and half tone images are selected arbitrarily. The original secret image can be extracted. First, the original two value secret images are shared with the N amplitude meaningless sharing image by Shamir sharing. Then, the self hiding technology of the halftone image will be used to share the image. The image is hidden in the half tone image to form n amplitude half tone sharing images. Third, from the N amplitude and half tone sharing image, the shared images hidden in n amplitude are extracted by the self extraction technology of halftone image. Finally, from the shared images extracted from the N amplitude, the K amplitude sharing image is chosen arbitrarily, and the hidden secret is extracted by Lagrange interpolation technique. (4) transform domain stereo video blind watermarking algorithm. On the basis of realizing and analyzing the performance based on DCT and DFT based video watermarking algorithm, a transform domain stereo video blind watermarking algorithm is proposed. The watermark and the DFT watermark which have the opposite embedding rules are embedded in the left and right image of the stereo video. After the video is attacked, The watermark is extracted from the relative relationship between the left and right path images, and the more robust watermark is extracted from the DCT watermark and the DFT watermark. (5) the efficient improvement of the image watermarking algorithm based on the DFT. On the basis of the realization and analysis of the image watermarking algorithm based on the DFT, an efficient image watermarking algorithm based on DFT is proposed. The watermark information is coded to ensure that the image is embedded in the watermark, and the correct rate is reached to 100% under the condition of the attack. For each 4 x 4 frequency domain data block, the watermark is changed by the right upper left and the left to right, and the energy relationship is expressed with the 2 x 1 frequency domain points. Under the premise that PSNR is basically the same, the embedding capacity is increased by 2 times. The watermark capacity is expanded by adjusting the range of embedding watermark.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41;TP309
[Abstract]:With the continuous development of computer and network technology, while people enjoy a lot of resources on the Internet, how to ensure the security of information has become an urgent problem. Information hiding is a major technology in the field of information security. Image and video information hiding is the most commonly used form, generally refers to a specific way to the image. The data of video and video are modified to hide information. After hiding, the use of images and videos is not affected. The halftone image is a special image carrier, only by two color values of black and white to represent an image. The information hiding in the halftone image can be divided into two categories: (1) a dot watermark: a halftone image through a dot. Features, direct hidden watermark information; (2) halftone visual encryption (halftone visual cryptography, HVC): in multiple halftone images, hidden a secret pattern. In HVC, (2,2) HVC is most commonly used, a secret pattern is hidden in two halftone images, and the secret pattern can be extracted by processing two halftone images. The former (2,2) HVC algorithm needs to be improved in extracting the correctness of the pattern (CDR) and generating the quality of the image. In addition, the existing (k, n) HVC algorithm needs codebook and will cause pixel expansion, that is, with the increase of the number of users n, the resolution of the generated half tone image needs to be extended to many times the original image. The stereo video is a kind of one. Special video carrier, each frame video includes images of about two channels. The images of the two channels are similar, but different. Stereo video watermarking becomes a new research direction of video watermarking. Because of the actual needs of the project, we hope to achieve a kind of image watermarking algorithm that can resist many attacks, including: coding attack, cutting attack. Attack, straight line attack, zoom attack, and can guarantee a certain watermark capacity. Aiming at all the problems above, this paper focuses on halftone information hiding, stereo video watermarking and image watermarking. The main achievements are as follows: (1) a new conjugate based color (2,2) halftone visual encryption algorithm. A new method for color halftone images is proposed. A new conjugate data hiding algorithm based on error diffusion (Data hiding by new conjugate color error diffusion, abbreviated NCCED) and a new conjugate data hiding algorithm based on point diffusion (Data hiding by) are applied to the concept of conjugation, which is applied to error diffusion half tone image and point diffusion half tone image information hiding. Conjugate color dot diffusion, abbreviated as NCCDD). The two algorithms hide the secret pattern in two color half tone images. By executing these two algorithms, the number of components needed to be replaced in the process of color half tone image information hiding, and the situation that the data hiding can not be implemented because the displacement interference is too large and can be improved. The correct decoding rate (CDR) of the secret pattern can also be extracted, and the quality of the generated half tone image can be improved; the concept of average per pixel interference value (Average distortion per pixel, abbreviated for short) is proposed, and the influence of the information hiding algorithm on the quality of the generated halftone image is measured by the size of ADPP. (2) high quality (2,2) halftone based on block conjugation According to the encryption algorithm, multiple half tone images are selected as training samples, and the minimum mean square filter (Least mean square filter, called LMSF) is generated to evaluate the quality of the halftone image with LMSF. In order to improve the quality of the half tone image generated by the information hiding algorithm, the concept of block conjugation is proposed and the block conjugation concept is applied to the halftone information hiding. A high quality halftone information hiding algorithm based on block conjugation (High quality data hiding in halftone images based on block conjugate, referred to as HQDHBC) is proposed. The algorithm hides a secret pattern in two halftone images based on error diffusion. When the conjugate or the same, the last point does not affect the block conjugate relation, the last point executes the normal halftone method and does not carry out the information hiding algorithm. Because of the execution of the HQDHBC algorithm, the number of information hiding algorithms can be reduced without affecting the accuracy of the secret pattern extraction, thus improving the quality of the generated halftone image. (3) the (k, n) half tone visual encryption algorithm based on Shamir sharing. The Shamir secret sharing is applied to the halftone visual encryption, and a K, n semi tone visual encryption algorithm based on Shamir secret sharing is proposed. Halftone visual encryption, a two value secret image is hidden in the N amplitude and half tone image, and K amplitude and half tone images are selected arbitrarily. The original secret image can be extracted. First, the original two value secret images are shared with the N amplitude meaningless sharing image by Shamir sharing. Then, the self hiding technology of the halftone image will be used to share the image. The image is hidden in the half tone image to form n amplitude half tone sharing images. Third, from the N amplitude and half tone sharing image, the shared images hidden in n amplitude are extracted by the self extraction technology of halftone image. Finally, from the shared images extracted from the N amplitude, the K amplitude sharing image is chosen arbitrarily, and the hidden secret is extracted by Lagrange interpolation technique. (4) transform domain stereo video blind watermarking algorithm. On the basis of realizing and analyzing the performance based on DCT and DFT based video watermarking algorithm, a transform domain stereo video blind watermarking algorithm is proposed. The watermark and the DFT watermark which have the opposite embedding rules are embedded in the left and right image of the stereo video. After the video is attacked, The watermark is extracted from the relative relationship between the left and right path images, and the more robust watermark is extracted from the DCT watermark and the DFT watermark. (5) the efficient improvement of the image watermarking algorithm based on the DFT. On the basis of the realization and analysis of the image watermarking algorithm based on the DFT, an efficient image watermarking algorithm based on DFT is proposed. The watermark information is coded to ensure that the image is embedded in the watermark, and the correct rate is reached to 100% under the condition of the attack. For each 4 x 4 frequency domain data block, the watermark is changed by the right upper left and the left to right, and the energy relationship is expressed with the 2 x 1 frequency domain points. Under the premise that PSNR is basically the same, the embedding capacity is increased by 2 times. The watermark capacity is expanded by adjusting the range of embedding watermark.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41;TP309
【参考文献】
相关期刊论文 前10条
1 刘U啿，
本文编号：2129737
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2129737.html
