基于纹理特征的高保真可逆水印研究及其在医学上的应用

发布时间：2018-04-27 21:49

本文选题：可逆水印 + 图像纹理　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：随着数字化医学影像技术的不断发展,不同医疗机构之间的信息交流也随之愈加方便快捷,但与此同时如何对数字化医学影像中所包含的患者私密信息加以保护,也日益受到人们的关注。医学领域的信息敏感性也使得任何对原始图像的改变都是不被允许的,进一步增大了医学图像私密信息保护的难度。可逆水印技术作为数字水印技术的重要分支,可以将患者的私密信息隐藏在医学图像中,并在提取水印的同时无损恢复原始载体图像,在医学图像等信息敏感领域具有较广泛的应用。传统的可逆水印算法主要通过利用图像的平滑区域以取得较好的嵌入效果,而医学图像往往具有丰富的纹理区域,因此有必要对其局部纹理性加以有效地利用。本文主要研究通过对医学图像局部纹理信息的有效利用,改进已有的可逆水印方法的嵌入效果。本文在对预测误差扩展技术的可逆水印方法的基本原理、主要过程和失真分析加以充分介绍的基础上,结合边缘检测原理和负载分配策略分别提出了两种基于医学图像局部纹理性的可逆水印改进方法。主要工作如下:通过对图像边缘检测的相关原理加以分析,本文提出了一种通过对图像局部区域纹理信息加以归类,进而用不同预测模型进行像素灰度值预测的预测方法。通过使用所提出的预测方法结合误差扩展技术提出了一种新的可逆水印方法。通过与已有可逆水印方法进行对比实验,验证了所提出方法在预测精度、嵌入容量和嵌入失真控制方面均取得了较好的改进效果。通过对彩色医学图像不同颜色通道之间相关性的研究,提出了一种利用色彩通道间纹理信息相似性来改进待嵌入序列排序策略的方法。并在此基础上改进了已有的全局遍历搜索的水印负载分配方法,使用贪心策略以构建的方式代替全局搜索寻找负载分配组合,使得嵌入效果与最优分配方案近似的同时极大改善了其时间效率。
[Abstract]:With the development of digital medical imaging technology, information exchange among different medical institutions becomes more and more convenient, but at the same time, how to protect the private information of patients contained in digital medical images, Also get people's attention day by day. The information sensitivity in medical field also makes any changes to the original image not allowed, which makes it more difficult to protect the private information of medical image. As an important branch of digital watermarking, reversible watermarking can hide the private information of patients in medical images, and recover the original image without loss while extracting the watermark. It is widely used in information sensitive fields such as medical images. The traditional reversible watermarking algorithm mainly uses the smooth region of the image to get better embedding effect, but the medical image often has rich texture area, so it is necessary to make effective use of the local texture of the image. This paper focuses on improving the embedding effect of existing reversible watermarking methods by using the local texture information of medical images effectively. In this paper, the basic principle, main process and distortion analysis of reversible watermarking based on prediction error expansion technique are introduced. Combined with the principle of edge detection and load allocation strategy, two reversible watermarking methods based on local texture rationality of medical image are proposed. The main work is as follows: by analyzing the relevant principles of image edge detection, this paper proposes a prediction method of pixel gray value by classifying the texture information of the local region of the image, and then using different prediction models. A new reversible watermarking method is proposed by combining the proposed prediction method with error expansion technique. Compared with the existing reversible watermarking methods, the proposed method has achieved better results in prediction accuracy, embedding capacity and embedding distortion control. Based on the study of the correlation between different color channels of color medical images, a method is proposed to improve the sorting strategy of the sequence to be embedded by using the similarity of texture information between color channels. On this basis, the existing load allocation method of global traversal search is improved, and the greedy strategy is used to replace the global search to find the load allocation combination. The embedding effect is similar to the optimal allocation scheme and its time efficiency is greatly improved.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R319;TP309.7

【相似文献】