基于LBP人脸纹理特征的差分直方图移位无损信息隐藏算法  

Differential histogram shift lossless information hiding algorithm based on LBP face texture features

作　　者：张弢[1] 柳雨农 任帅[2] 张德刚 Zhang Tao;Liu Yunong;Ren Shuai;Zhang Degang(School of Electronic&Control Engineering,Chang’an University,Xi’an 710068,China;School of Information Engineering,Chang’an University,Xi’an 710068,China;Education Training Evaluation Center,Yunnan Power Grid Co.Ltd.,Kunming 650033,China)

机构地区：[1]长安大学电子与控制工程学院,西安710068 [2]长安大学信息工程学院,西安710068 [3]云南电网有限责任公司教育培训评价中心,昆明650033

出　　处：《计算机应用研究》2020年第6期1774-1778,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61702050,61402052);国家级大学生创新创业训练计划资助项目(201610710036);中央高校“双一流”引导专项资助项目(300104292405);长安大学中央高校基地科研业务费专项资金资助项目(300102240208);2020年长安大学研究生教育教学改革资助项目(300103102417);2019年长安大学研究生教育教学改革建设专项基础建设项目(300103190640);大学生创新创业训练计划资助项目(201910710079,201810710052,201810710060,201810710215,201810710224)。

摘　　要：针对信息隐藏算法中提高嵌入量与增强鲁棒性之间的矛盾问题,提出一种多载体信息隐藏算法。使用多幅人脸表情图像作为载体,采用局部二值模式(LBP)纹理特征识别人脸表情区域来嵌入加密信息;计算出载体区域的相邻像素差值矩阵,通过对差值矩阵的对应元素直方图进行移位构造出嵌入空间来实现加密信息的可逆隐藏与载体图像的无损恢复。分析证明了该算法比现有算法具有更大的嵌入容量并保持较高的鲁棒性的优势,在最大嵌入容量达到0. 561时具有38. 421 d B的信噪比(PSNR),且在识别的嵌入区域PSNR值达到46. 286。鲁棒性实验表明,该算法对于滤波攻击可以与原始信息图像的相似度大于99%;面对剪切、平移攻击时,秘密图像归一化系数NC最小为0. 743和0. 728,远大于其他算法。从与其他算法的对比实验结果看,该算法是有效的。Aiming at the contradiction between improving the embedded quantity and enhancing the robustness in the information hiding algorithm,this paper proposed a multi-carrier information hiding algorithm. The algorithm used multiple facial expression images as vectors,and applied local binary pattern( LBP) texture features to recognize facial expression regions to embed encrypted information. It calculated the adjacent pixel difference matrix of the carrier region,and constructed the embedded space by shifting the histogram of the corresponding element of the difference matrix,in order to realize the reversible concealment of the encrypted information and the lossless recovery of the carrier images. Algorithm analysis proves that it has greater embedding capacity and maintains higher robustness than existing algorithms. Its largest embedded capacity reaches0. 561,at same time the peak signal to noise ratio( PSNR) is 38. 421 d B. And the value of embedded region PSNR reaches46. 286. The robustness experiments show that the proposed algorithm may has a similarity greater than 99% to the original information image for the filtering attack. In the face of shear and translation attacks,the minimum image normalization coefficient is 0. 743 and 0. 728,which is much larger than other algorithms. From the comparison experimental results with other algorithms,the proposed algorithm is effective.

关 键 词：无损信息隐藏 多载体 局部二值模式 差分矩阵 直方图移位 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]