基于人眼对比度感知特性的数字图像防伪技术研究  

Anti-counterfeiting Technology of Digital Image Based on Human Vision System

作　　者：姚军财[1] 

机构地区：[1]陕西理工学院物理与电信工程学院,陕西汉中723000

出　　处：《光电子技术》2011年第4期226-232,239,共8页Optoelectronic Technology

基　　金：国家自然科学基金项目(10447005);陕西省教育厅专项科研基金资助项目(2010JK464)

摘　　要：结合人眼视觉特性和数字图像傅里叶变换频谱图的特点,提出了一种新的人眼对图像感知的最小可觉察误差即JND阈值的计算方法,根据计算的JND阈值提出了一种基于人眼对比度敏感视觉特性的数字图像水印算法,并加以实现和攻击测试;实验结果得出:经过不同质量因子QF的JPEG压缩,峰值信噪比在QF大于40%时均大于30 dB,提取的水印图的相似度NC在QF=20%时仍能达到0.923 9,表明水印信息具有较好的抗压缩攻击性;且因为引入了人眼视觉特性,使得含水印图具有较好的透明性;由于在嵌入过程中引入了JND阈值,嵌入算法使得水印嵌入的位置具有随机性和嵌入量达到了最大,从而有力地保证了水印信息的安全性。实验结果表明提出的水印方案是有效的、可行的。Based on human vision characteristics and frequency spectrum characteristics of digital image Fourier transform, a JND threshold calculating method was put forward with respect to tile threshold, above which the human eye can detect noticeable difference of image. A digital watermark algorithm was proposed according to the JND threshold and the result showed： PSNR was larger than 30dB when quality factor was larger than than 40%, ;NC could reach 0. 9239 through JPEG compression of different quality factor,when quality factor was 20%, this indicated that the watermark had good resistance to compression attack. The watermarked image had good transparency because of the introduction of the human visual characteristics. Owing to JND threshold＇ s security improvement, the embedded location in the embedding process was random and embedded quantity reached the maximum, this showed that the proposed method is feasible and effective.

关 键 词：数字图像 对比度敏感函数 水印 JND模型 

分 类 号：TN29[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]