查询结果:   李爱华,刘金华.基于Gauss-Hermite分布模型的水印检测算法研究[J].计算机应用与软件,2017,34(9):216 - 221,266.
中文标题
基于Gauss-Hermite分布模型的水印检测算法研究
发表栏目
图像处理与应用
摘要点击数
663
英文标题
A WATERMARK DETECTION ALGORITHM BASED ON GAUSS-HERMITE DISTRIBUTION MODEL
作 者
李爱华 刘金华 Li Aihua Liu Jinhua
作者单位
南昌大学抚州医学院 江西 抚州 344000 电子科技大学电子工程学院 四川 成都 611731    
英文单位
Fuzhou Medical College of Nanchang University,Fuzhou 344000,Jiangxi,China School of Electronic Engineering,University of Electronic Science and Technology of China,Chengdu 611731,Sichuan,China    
关键词
水印检测方法 高斯-埃尔米特分布模型 视觉感知 工作特性曲线
Keywords
Watermark detection method Gauss-Hermite distribution model Visual perceptual Receiver operating characteristic
基金项目
作者资料
李爱华,讲师,主研领域:图像处理。刘金华,高工。 。
文章摘要
图像水印的检测来源于统计信号处理的思想,其关键在于如何对图像经变换后的系数进行建模。在已有的水印检测模型中,图像基于高斯分布进行建模,容易导致在抗攻击条件下水印检测的性能较低。为克服这一不足,结合图像的视觉感知模型和高斯-埃尔米特(Gauss-Hermite)统计分布模型,提出一种数字水印的嵌入和盲检测方法。在水印嵌入中,以临界可见误差值JND(Just Noticeable Difference)作为水印嵌入强度,以提高水印的不可感知性。在水印检测中,基于高斯-埃尔米特分布对图像小波系数进行建模,并通过假设检验提出了水印的盲检测模型,最后推导了虚警概率和检测概率之间的工作特性关系。仿真结果表明:提出的水印检测模型在诸如噪声、JPEG压缩、滤波以及缩放等攻击下具有较好的性能。
Abstract
The detection of image watermarking comes from the idea of statistical signal processing. Its key point is how to model the transformation coefficient of an image. In the existing detection model, the coefficient of image is modeled by Gauss distribution, which results in the low performance of watermark detection under attack. In order to address this issue, a watermark embedding and blind detection method was proposed by combining the visual perceptual model and Gauss-Hermite distribution model. To improve the invisibility of watermarking, Just Noticeable Difference was used as the watermark embedding strength factor during the embedding process. In the process of watermark detection, the image wavelet coefficients were modeled by Gauss-Hermite model, thus the blind watermark detection method was proposed. Finally, the Receiver Operating Characteristic (ROC) curve between the probability of false alarm and the probability of detection was derived and discussed respectively. Experimental results demonstrate that the proposed method has better performance under attack such as AWGN, JPEG compression, filtering and scaling.
下载PDF全文