基于HVS的半色调图像数字水印算法研究外文翻译资料

 2023-04-13 10:32:39

A Study of Digital Watermark Algorithm Based on HVS for Halftone Images

Feng Chunhui, Liu Quanxiang

Keywords: Halftone; Digital watermark; Human Visual System; Amplitude Modulation; Texture

Abstract. Special printing material and special techniques are used for the purpose of fighting against the piracy of the presswork. But these methods require high standard technology and high production costs. Using digital watermark can guarantee a strong and low cost protection for the presswork. Existing digital watermark algorithms for halftone images are mostly applied to FM (Frequency Modulation) screening and these algorithms usually complete the embedding process by modulating the coefficients of the frequency domain. Such algorithms are relatively complex, and they are easy to lose information so that it will make a lower detection probability. Moreover, in the actual printing process, AM (Amplitude Modulation) screening is the more mature and more widely used method compared with FM screening. The subject of this article is monochrome presswork generated by AM screening. The structure of this article is as follows: The first section uses the principle of HVS (Human Visual System) model to select the area in which the watermark will be embedded. The second section uses different screening threshold matrixes to create the printing dots in the selected area. In this way, a texture of certain direction is formed. The last section uses Fourier Power Spectrum method to detect the texture features of the screened image to complete the detection of the watermark. The result of the experiment shows that this algorithm can generate a digital watermark of good transparency and high detection effectiveness. It combines a unique AM screening approach to complete the embedding process. All the processes do not require any additional equipment, and the embedding and detection process are quite simple.

What is the digital watermarking technology

Digital watermark technology is through a certain algorithm will landmarkinformation directly embedded into the multimedia content, but not affect theoriginal content value and use, and can not be aware of the perceptual system or pay attention to, only through a special detector or reader to extract.Digital watermarking technique can distinguish whether the object is under protection, monitoring of the protected data transmission, authentication andillegal copy, resolve copyright disputes,and to provide evidence to the court.Copyright protection digital watermarking to requirements.

1)validity: in copyright protection digital watermarking embedded content should have authority, that is to say should be able to show that the carrierfile ownership is recognized;2) a robust watermarking anti-attack ability; 3) capacity: refers to embed information in the file number; 4)imperceptibility:refers to the embedded watermark information file with the original documents should be almost the same or does not influence the normaluse of the files, but also users can not feel the existence of watermark.

Digital watermarking technology in application of electronic reading

Digital watermarking of a series of technical characteristics show that, in therapid development of the electronic reading service, digital watermarking willbe widely used.For image, video watermarking algorithm Discrete cosine transform digital watermarking is often used as a basictransformation.First introduced the definition of discrete cosine transform:On digital image,Two dimensional inverse DCT transform ( IDCT ) is defined as:By 3.2 the original image can be expressed as, for a series of weights

When a pair of pixels in an image matrix after discrete cosine transform, thefrequency domain matrix in left corner element value maximum, for the DCcomponent, representing the whole image of the average brightness; whilethe remainder of each element value according to the corner element forfixed-point triangle lateral snaking arranged successively, representing the image low frequency, medium frequency components and high frequencycomponents.

Cox presents a global DCT domain watermarking algorithm:

Selection of sequences of times; = x1, x2,... , xn as watermark, where Xi is satisfythe Gauss distribution of N (0,1) of a random number.The algorithm first uses the discrete cosine transform of the original image l is transformed intothe frequency domain, using the D representation of the data obtained.Fromthe coefficient of D select n most important frequency component, consistingof sequences of V = V1, v2,... ,VN, in order to improve the robustness ofimage compression.

Get the watermark array and frequency domain array, using the formula: Vi=vi(1 a Xl ) digital watermark is embedded into the frequency domain array carrying watermark information, get new frequency domain array sequence ofv = V1, v2,...,VN, then the V' of each element in the frequency domain matrix D in the corresponding position of the V element replacement out, get anew frequency domain matrix D, the D inverse discrete cosine transform becontaining the watermark watermark image l. In the watermark image l* forwatermark detection, the whole process is the reverse process of embeddingprocess almost.The first is to l* through DCT to obtain the frequency domainarray sequence of V* = v1*, v2*.... , vn*, then according to the formula Vi=vi(1 a Xl ) inverse formula one watermark sequence of X*= x1*,x2*.... , xn*,then calculate the watermark with the original watermark correlation of X*X.

The algorithm is robust, can resist including scaling,JPEG compression,shear and jitter, printing and copy - scann

剩余内容已隐藏,支付完成后下载完整资料


基于HVS的半色调图像数字水印算法研究

摘要:使用特殊的印刷材料和特殊的印刷技术打击盗版。但这些方法要求技术标准高,生产成本高。采用数字水印可以保证对印刷品的保护力度大、成本低。现有的半色调图像数字水印算法多应用于FM (Frequency Modulation)滤波,这些算法通常通过调制频域系数来完成嵌入过程。这类算法比较复杂,容易丢失信息,检测概率较低。另外,在实际印刷过程中,AM(调幅)增网是比FM增网更成熟、应用更广泛的方法。本文的主题是单色印刷由AM筛分产生。本文的结构如下:第一部分利用HVS (Human Visual System)模型的原理选择水印要嵌入的区域。第二部分使用不同的筛选阈值矩阵在选定区域中创建打印点。这样就形成了具有一定方向的纹理。最后一节采用傅里叶功率谱法检测被筛选图像的纹理特征,完成水印的检测。实验结果表明,该算法能生成透明度好、检测效率高的数字水印。它结合了独特的AM筛选方法来完成嵌入过程。所有工艺都不需要任何额外的设备,且埋设和检测过程相当简单。

什么是数字水印技术?

数字水印技术是通过一定的算法将地标性信息直接嵌入到多媒体内容中,而不影响原有内容的价值和使用,并不能感知其感性系统或注意事项时,只能通过专用探测器或读卡器来提取。数字水印技术可以区分目标是否被隐藏、保护、监控受保护数据的传输、认证和非法复制,解决版权纠纷,并向法院提供证据。

数字水印符合版权保护要求

1)有效性:数字水印在版权保护中嵌入内容应该有权限,也就是说应该能够证明运营商文件的所有权是被认可的;2):具有鲁棒水印抗攻击能力;3)容量:指在文件号中嵌入的信息;4)不可察觉性:指嵌入的水印信息与原始文件相同,文档应该几乎相同或不影响文件的正常使用,同时用户也感觉不到水印的存在。

数字水印技术在电子阅读中的应用

数字水印的一系列技术特点表明,在快速发展的电子阅读服务中,数字水印将得到广泛的应用。对于图像,视频水印算法离散余弦变换数字水印常被用作一种基本变换。二维反DCT变换(IDCT)定义为:

当一对像素在图像矩阵中经过离散余弦变换后,频域矩阵中左角元素值最大,为DCcomponent,代表整个图像的平均亮度;而每个元素值的余数按定点三角形横向蛇行角元依次排列,表示图像的低频、中频和高频分量。

Cox提出了一种全局DCT域水印算法:

times; = x1, x2,hellip;, xn作为水印,其中Xi满足随机数N(0,1)的高斯分布。该算法首先利用离散余弦变换将原始图像l变换到频域,利用D表示所获得的数据。从D的系数中选取n个最重要的频率分量,由V = V1, v2,hellip;,以提高图像压缩的鲁棒性。

得到水印数组和频域数组,利用公式:Vi= Vi (1 a Xl)将数字水印嵌入到频域数组中携带水印信息,得到新的频域阵列序列v = V1, v2,hellip;,VN,则频域中每个元素的V”,将矩阵D在相应位置的V元素替换出来,得到新的频域矩阵D,将D逆离散余弦变换包含水印图像l。在水印图像l*中进行水印检测,整个过程几乎是嵌入过程的逆过程。第一种是l*通过DCT得到V* = v1*, v2*....的频域阵列序列, vn*,则根据公式Vi= Vi (1 a Xl)逆公式1水印序列X*= x1*,x2*...., xn*,然后利用原始水印相关性X*X计算水印。

该算法具有鲁棒性,能够抵抗包括缩放、JPEG压缩、剪切和抖动、打印和复制扫描、图像处理和嵌入等在内的多重水印,也能够抵抗多用户攻击,是一种非常有效的水印算法。

文本数字水印是什么

与图片或视频文件、文本文件没有太多的冗余,因此基于文本的水印技术远远小于图像或视频的水印技术。常用的文本水印主要是基于固定格式的文本水印。常用方法:基于间距信息的表示,基于字间距信息的表示方法和基于字符属性信息的表示方法表示。在实际使用中,这三种信息隐藏方法可以结合使用。

基于固定格式的文本水印,其最大的缺点是对保留文本而改变文本格式的软拷贝敏感。为了解决这一问题,人们提出了基于文本的语义水印技术,但该技术难以实现,且水印提取需要提供源文件控制,可操作性不强。

在实际的电子阅读中,我们将能够分发的内容使用格式开发者提供的工具进行内容集不可复制(word、PDF具有此功能),避免非法用户销毁水印信息。此外,还可以将文本扫描为图片格式,利用离散余弦变换技术嵌入水印,然后提供给用户。

总结

在电子阅读业务应用中,数字水印技术主要侧重于侵权行为的识别,也不能预防和制止侵权行为的发生。而数字水印算法的鲁棒性还不能完全满足要求。但随着数字水印算法的不断进步,以及catech技术和加密技术的使用,数字水印技术在电子阅读领域将有更广泛的应用。

介绍

现有的半色调图像数字水印算法多集中在调频滤波中,通常通过对频域系数进行调制来实现水印的嵌入。这种算法比较复杂,而且在印刷过程中,由像素构成的数字图像转换为点阵构成的半色调图像时,会对图像的高频系数产生较大的干扰,因此水印的鲁棒性很低。而AM筛选技术较为成熟,在筛选技术中占有主导地位。但是目前关于数字水印算法用于AM筛选的文献很少。本文的目的是使AM筛分印刷的数字水印技术能够实际应用,增强水印的鲁棒性和透明度,提高水印的嵌入检测速度和检测效果。

实验原理

基于DCT的Watson视觉模型。Watson模型[1]使每个8 times; 8像素的图像块经过离散余弦变换(DCT)得到DCT系数。低频系数变化较小时会出现视觉差异,高频系数变化较大时也会出现视觉差异。此外,根据沃森的理论,如果一个块的平均灰度较高,则相应的DCT系数的变化可能更大而不被人眼所察觉。本文使用平滑值代替频率值,这是因为Watson理论不适用于空间域;将水印嵌入到平滑度值和灰度值高于相应阈值的块中,以达到增强水印透明度的目的。

半色调图像的AM筛分理论。本文使用生长模型[2]生成打印点。假设每个印刷网格在水平方向和垂直方向上平均分为5个部分,形成25个曝光点。在这种情况下,理论上一个网格可以显示26个灰度级,即一个额外的曝光点产生一个额外的灰度级。但为了简单起见,我们只设置了14个灰度级,灰度级从0到13,网格将按照上面提到的方式进行划分,这样,每两个曝光点就可以表示为一个灰度级。(见标签。1)

表1灰度级与曝光点数的关系

灰度级

13

12

11

10

9

8

7

6

5

4

3

2

1

0

曝光点的数量

0

1

3

5

7

9

11

13

15

17

19

21

23

25

当需要再现一幅8位深的单色图像时,理论上需要16 times; 16 = 256个曝光点形成一个网格,才能再现图像的所有渐变;但由于人眼无法分辨一百或一百以上的[3]阶次,人们一般采用10 times; 10或12 times; 12的束带法。

水印嵌入的原理。在需要嵌入水印区域的奇偶网格行中分别使用两个不同的滤波阈值矩阵。例如,在奇数行中使用阈值矩阵A生成点,在偶数行中使用阈值矩阵b生成打印点,这样就形成了某一方向的纹理。理论上,在纹理区域,通过傅里叶功率谱分析,可以在能量角曲线中检测到明显的能量峰。

水印检测的原理。水印的检测就是对纹理特征的检测。本文采用傅立叶功率谱[4]方法检测图像的纹理特征。设纹理图像为f (x, y), f (x, y)表示在坐标(x, y)上的像素值,f (x, y)[5]的傅里叶变换可以表示为公式(1):

F (u, v) =int;int;F (x, y) exp{minus;j2pi;(用户体验 v)} .

  1. 如果傅里叶变换用极坐标表示,那么它就是F(r,theta;)的形式。通过了解一个小扇形区域在一定角度的能量,可以得出该扇形区域的能量变化是由该扇形区域所处的角度决定的。见公式(2)。Phi;theta;=int;F博士(r,theta;)2 .

(2)如果沿theta;方向存在大量的线和边,则在频域内,能量将沿方向集中theta; Pi;/2

实验方案

水印嵌入区域的选择。将一幅数字图像均匀分成若干块,每块为64 times; 64像素;分别计算每个块的平滑度值和灰度值,目的是选择更平滑、灰度值更高的块

灰度。但如何确定每个参数的阈值,就必须借助一些实验。平滑度值和灰度值[6]的公式:

R = 1 minus; . (3) sigma;2 = (4)

R为平滑度值;Zi是第i个像素的灰度值;M为整幅图像灰度的平均值;sigma;2是方差。

设某块的平均灰度为D,设整幅图像的平均灰度为D0,计算平滑度值R和灰度值V,设置R和V的阈值,如果某个块的R和V都大于其阈值,则选择该块作为嵌入水印的区域。K由所有被选中块的位置信息组成。

水印的嵌入。在对AM筛分原理进行讨论的基础上,采用了每个网格除以12 times; 12的方法。通过这种方式,每个网格可以复制144个灰度级。筛选阈值矩阵A如图2所示;将矩阵A向右平移得到阈值矩阵B。见图3。

图1筛选阈值矩阵 图2筛选阈值矩阵B

对于需要嵌入水印的块中网格的奇数行,采用矩阵a作为水印的筛选矩阵;偶行用矩阵B;在没有水印的区域使用矩阵A。经过这些步骤,得到了嵌入水印的数字点图像。

水印的检测。数字点图像生成后,也会按照分割数字像素图像的方法将数字点图像分割成块。利用傅立叶功率谱计算数字网点图像,生成每个块的能量角曲线。

对每条曲线进行分析,看是否有明显的峰值。如果是,则说明对应角度上存在纹理。通过对纹理来源的分析,可以判断使用不同的筛选阈值矩阵是否形成了纹理。

通过判断块的曲线特征确定为嵌入区域的块的位置信息设为K 。使用K 和K进行相似性比较,如果相似性大于某一阈值,则为成功的实验。

实验程序

水印嵌入区域的选择。实验是在一个特定的Visual c 程序上进行的。该程序可以将图像分块,并计算每个块的平滑度值和灰度值。根据程序输入的平滑度值和灰度值的阈值(见图4),选出两个参数都大于阈值的块。选中的方块如图5所示为灰色。

图3程序 图4所选块的位置

实验结果表明,平滑度阈值为0.05,灰度值阈值为1.22。将分辨率为350dpi的灰度图像读入程序,进入两个阈值后,生成水印位置的图形表示。因此,这些块的位置信息K(表2),根据图6得到。图6中,字母i表示每行序号,字母j表示每列序号。

图5始位置信息 图6输入阈值 图7带水印的点图像

表2嵌入水印的块的位置信息(K)

lt;

剩余内容已隐藏,支付完成后下载完整资料


资料编号:[589584],资料为PDF文档或Word文档,PDF文档可免费转换为Word

i

2

2

5

5

6

6

7

7

8

8

9

9

9

10

10

10

10

11

11

11

12

12

13

13

15

j

2

3

7

8

7

9

1

2

2

9

1

2

11

2

5

8

9

您需要先支付 30元 才能查看全部内容!立即支付

课题毕业论文、外文翻译、任务书、文献综述、开题报告、程序设计、图纸设计等资料可联系客服协助查找。