基于纹理分析的可逆图像水印算法研究

发布时间：2018-05-11 01:42

本文选题：可逆图像水印 + 纹理分析　；参考：《江西理工大学》2017年硕士论文

【摘要】：大家的学习和工作因为互联网的飞快发展给带来了极大的方便,同时也带来诸如盗版、信息篡改等一系列潜在的信息安全问题。为了解决该问题,传统的方法采用加密和数字签名等技术,但加密使得不法分子容易看出要传输的通信信息,从而获取和破解他们感兴趣的信息。而数字签名虽然能够为数据的传输提供有效的保护,但需要在原始数据中加入大量签名,同时随着并行计算的发展,数字签名的安全性已经受到质疑。为了解决传统技术上的缺陷问题,数字水印技术被提出,利用载体冗余性来嵌入隐蔽信息,提高载体传输的安全性。但是传统的数字图像水印在嵌入隐藏信息后,载体图像无法还原到最初始状态,对图像质量要求极高的特殊范围领域是难以接受的。为了解决该问题,学者们提出了可逆水印,可逆水印能够无损恢复最初始图像和完全提取隐藏信息,在医学、工程、军事等领域具有很好的发展前景。本文以灰度图像为研究对象,通过图像的纹理特征分析,提出基于纹理分析的无损水印算法,本文的主要工作如下:(1)针对预测器的预测准确度和隐藏信息图像遮蔽性上不足的问题,提出利用领域八个方向的梯度预测和自适应选择水印嵌入块的可逆图像水印。该方法充分考虑到图像周围像素的相关性,利用均方误差给邻域八个像素分配权值,然后计算预测像素值。并采用均方误差分析每个子块纹理复杂度,选择纹理最复杂的子块进行水印嵌入,在嵌入区域采用基于预测误差对方法嵌入水印。算法提高了水印的遮蔽性和预测精准度。(2)针对灰度共生矩阵的四个主要特征参数与图像子块纹理复杂度之间没有明显的数学关系和水印遮蔽性上不足的问题,提出一种基于灰度共生矩阵特征参数用于分析图像纹理的可逆图像水印。通过提取灰度共生矩阵四个特有的特征参数,利用均方误差给每个特征参数分配权重值,从而建立灰度共生矩阵的特征参数与载体图像纹理复杂度的数学关系;然后把宿主图像分块,利用图像复杂度的数学关系去计算子块纹理复杂度,选择复杂度最大的子块去嵌入隐藏信息。该方法对于自然图像和医学图像尤其是纹理复杂的图像取得更好的嵌入效果。
[Abstract]:Because of the rapid development of the Internet, people's study and work bring great convenience, but also bring about a series of potential information security problems, such as piracy, information tampering and so on. In order to solve this problem, the traditional methods use encryption and digital signature techniques, but encryption makes it easy for illegal elements to see the communication information to be transmitted, so as to obtain and decipher the information they are interested in. Although digital signature can provide effective protection for data transmission, it needs to add a large number of signatures to the original data. With the development of parallel computing, the security of digital signature has been questioned. In order to solve the problem of traditional technology, digital watermarking technology is proposed, which uses carrier redundancy to embed hidden information to improve the security of carrier transmission. However, the traditional digital image watermarking can not be restored to the initial state after embedding hidden information, so it is difficult to accept the special field which requires high image quality. In order to solve this problem, researchers have proposed reversible watermarking, which can restore the initial image without damage and extract hidden information completely. It has a good prospect in medicine, engineering, military and other fields. In this paper, a lossless watermarking algorithm based on texture analysis is proposed by analyzing the texture features of gray-scale images. The main work of this paper is as follows: (1) aiming at the problem of poor prediction accuracy and hidden information image masking of the predictor, this paper proposes to use the gradient prediction of eight directions of the domain and the adaptive selection of reversible image watermarking in the watermark embedding block. In this method, the correlation of the pixels around the image is fully taken into account, and the mean square error is used to distribute the weights of the neighboring eight pixels, and then calculate the predicted pixel values. The texture complexity of each sub-block is analyzed by mean square error, and the most complex texture sub-block is selected to embed the watermark, and the prediction error pair method is used to embed the watermark in the embedded region. The algorithm improves watermark masking and prediction accuracy. (2) there is no obvious mathematical relation between the four main characteristic parameters of gray level co-occurrence matrix and texture complexity of image subblock, and the problem of watermark masking is insufficient. A reversible image watermarking based on gray level co-occurrence matrix feature parameters is proposed to analyze image texture. By extracting the four characteristic parameters of the gray level co-occurrence matrix and using the mean square error to assign the weight value to each characteristic parameter, the mathematical relationship between the feature parameters of the gray level co-occurrence matrix and the texture complexity of the carrier image is established. Then the host image is divided into blocks, and the texture complexity of the sub-block is calculated by using the mathematical relation of image complexity, and the most complex sub-block is selected to embed the hidden information. This method can achieve better embedding effect for natural and medical images, especially for images with complicated texture.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309.7

【参考文献】