基于BJND的立体图像篡改定位及恢复水印方法

针对立体图像内容的真实性认证和完整性保护等 问题,提出一种基于双目恰可觉察失 真(BJND)的立体图像 篡改定位及恢复水印方法。首先,利用奇异值的稳定特性,设计左右图像的定位水印；然后 ,根据BJND强度将左右图像 的定位水印嵌入到左图像中；最后,利用离散余弦变换(DCT)和JPEG量化压缩 生成恢复信息,并将右图像和左图像遮挡暴露区 域的恢复信息分别嵌入到右图像和左图像中。实验结果表明,本文方法所构造的水印对JPEG 压缩、椒盐噪声、高斯白噪声等偶然攻击比较鲁棒,而对剪切、拼接、旋转等恶意攻击较为 脆弱,同时对左右图像的恶意篡改区域检测率大于98%。由于充分利用了左右图像的匹配特 性,恢复的被篡改立体图像其左右图像PSNR达36.02dB ～41.29dB。

BJND based watermarking method for tamper localization and recovery of stereoscopic image

To address issues of content authenticity verification and integrity protection, a watermarking method based on binocular just-noticeable-difference (BJND) is proposed for tamper localization and recovery of stereos copic images.Firstly,for its strong stability,singular values of the left and ri ght images of a stereoscopic image are used to construct authentication-bits fo r the left and right images.Then,the generated stereoscopic image authentication -bits are embedded into the left image by making use of the binocular just-not iceable-different value.Finally,the recovery information of right image and occ lusion parts in the left image is produced with discrete cosine transform and JP EG quantization compression,and embedded into the right and left images themselv es,respectively.The experimental results show that the watermarking constructed with the proposed method is robust to incident attacks,such as JPEG compression, salt and peppy noise and Gaussian white noise,but relatively fragile to mailicio us attacks,for example,crop,forgery,rotate,etc.The accuracy of tamper detection is higher than 98% for malicious attacks.For different stereoscopic images degra ded by malicious attakcs,the peak signal-to-noise ratios (PSN Rs) of the recovered left and right images of the stereoscopic im ages are from 36.02dB to 41.29dB,by taking full advantage of the matching char acteristics between the left and right images.