【作者】 刘芳；

【导师】 汪玉凯；

【作者基本信息】 大连理工大学 ， 信息管理与电子政务， 2013， 博士

【摘要】 可逆信息隐藏技术是近几年信息隐藏领域一个新的研究热点。不同于传统的信息隐藏技术在嵌入信息的同时会造成原始载体的永久失真,可逆信息隐藏技术能够在正确提取信息后无损地恢复原始载体,这在对载体信号的质量要求严苛的场合具有很高的应用价值。本文着重对基于图像的可逆信息隐藏技术进行了研究,主要贡献如下：1、提出了一种基于五像素预测误差扩展的可逆信息隐藏算法。算法以五个像素为一组来划分图像,在每个图像块中利用三对预测误差之间的关系嵌入秘密信息。最好的情况下,单个图像块中,每个像素的值最多改变1,即可嵌入六比特数据,有效地扩充了隐藏容量。同时,算法通过对图像块复杂度的识别,摒弃了纹理复杂的块,只选择纹理简单的块嵌入数据,降低了高容量对图像质量的损耗。2、提出了一种基于插值预测和块分类模型的可逆信息隐藏算法。对已有的直方图平移算法进行研究后发现,如果在不影响可逆性的同时,控制平移点的数量和减小秘密信息流中“1”的比例,算法的隐蔽性可以得到进一步提高。因此,本算法一方面通过设计一个图像块类型的预测器,同时对秘密信息进行预处理,限定了嵌入信息时被修改像素的数量,进而提高了隐密图像的质量。另一方面,对图像进行插值预测,采用插值误差直方图双向平移的方法隐藏信息,保证了较高的嵌入容量。3、提出了一种基于插值运算和隐写编码的二值图像可逆信息隐藏算法。算法重点解决了三个方面的问题：一是实现可逆。鉴于二值图像在实际存储时,每个像素占用一个比特,数据量较少,因此在嵌入提取信息时,对图像先扩后缩：嵌入信息时,对图像进行插值运算,在插值后的图像中寻找特定的插值点隐藏数据。提取信息后,通过重组非插值点来恢复原始图像,实现了算法的可逆性。二是减小图像失真。设定“可翻转点”选取规则,尽可能选取对人眼视觉影响较小的像素点来隐藏秘密信息。三是提升嵌入容量。应用隐写编码,在m×n的图像块中通过至多修改两个像素点,来隐藏r比特数据,r=log2(mn+1)。对于大小为M×N的插值后图像,其最大嵌入容量可达到M×N/m×n×r比特。4、从实际应用的角度出发,设计并实现了一个融合可逆信息隐藏技术和鲁棒水印技术的图像保护应用系统。系统实现了对图像的双重保护,并且通过标准的REST网络服务接口进行通讯,可以方便地集成到诸如远程医疗系统、电子政务系统等应用系统中。同时,针对鲁棒水印技术进行了研究,提出了一种抗抖动攻击的鲁棒水印算法。算法对图像进行DCT变换,以9个相邻块为一组划分单元,在每个单元内通过调整中心块的DC系数值嵌入和提取水印。实验结果显示,算法抗抖动等攻击的能力较强。更多还原

【Abstract】 Reversible data hiding is one of the hot research areas of information hiding in recent years. Unlike the traditional techniques, by which the source data is distorted permanently due to data hiding, the reversible techniques can fully restore the source data after the hidden data being correctly extracted. It is high valuable in the cases which distortion is intolerable. This paper mainly focuses on the image reversible data hiding algorithms, the major contributions are as the followings:1. A reversible data hiding algorithm based on five-pixel block and prediction-error expansion is proposed. The image is divided into five-pixel blocks, the secret data can be embedded by using the relationship between three pairs of prediction-error. In the best situation, each pixel value is modified at most by1in each block to embed six bits, in this way the high capacity is achieved. At the same time by identifying the complexity of image blocks, only the ones of simple texture are selected to hide data and the others of complexity are dropped, the impact on image quality is minimized.2. A reversible data hiding algorithm base on interpolation prediction and blocking model is proposed. The analysis on the previous histogram shifting algorithms shows that the quality of the stego-image can be improved by reducing the number of shifting points and the number of "1" in the secrete data without having impact on reversibility. Therefore a block predictor is designed and the secret data is pre-processed to improve the quality. Furthermore, the interpolation-error histogram is generated using interpolation prediction in the cover image, the histogram is shifted in two directions to have the high capacity.3. A binary image reversilbe data hiding algorithm using interpolation computation and steganographic coding is proposed. Three problems are resolved:the first is to have reversibility. Based on the fact that in binary image each pixel is saved in one bit and the required space is not big, the image is expanded when embedding data and is shrinked when extracting the data. The stego-image is obtained by hiding secret data in the particular pixels of the interpolated image; the cover image can be restored by re-arranging the non-interpolated pixels. The second is to decrease image distoration. Only the pixels with less effect on human being visual system can be selected to hide the secret data by setting the rules of modification. The third is to improve the capacity. By using steganographic coding, r bits data can be hidden in one block with changing up to two pixels, r=log2(mn+1). For the image with size of M×N, the maximum capactity can reach to (M×N)/(m×n)×r bits.4. In practice, a combined technical image protection system of reversible data hiding and robust watermarking is designed and implemented. The system implements two levels of protection for the cover image and can communicate with other application systems based on REST network interface. It can be easily integrated to the existing systems like remote-medical or E-government. At the same time, an anti-dithering attack algorithm is proposed after studying the robust watermarking. The image is divided into units with nine adjacent blocks in DCT domain, then the watermark is embedded and extracted by adjusting the DC coefficient of the center block. The experimental results show that the algorithm is robust for dithering and other attacks.更多还原