采用Hilbert扫描序列短重码统计的盲隐写检测方法  

Blind detection for image steganography using short duplicate codes statistical model for Hilbert scanning sequences

作　　者：钟尚平[1] 徐巧芬[1] 郭文忠[1] 廖彬[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350108

出　　处：《通信学报》2013年第1期51-60,共10页Journal on Communications

基　　金：国家自然科学基金资助项目(61103175);福建省自然科学基金资助项目(2010J01331);教育部科学技术研究重点基金资助项目(212086)~~

摘　　要：在LSB行扫描序列中,基于短重码间距统计的隐写检测方法对LSB匹配等隐写技术具有良好的盲检测性能。然而此方法只能利用相邻码元的相关性,影响检测性能。理论证明了短重码间距统计量的检测能力与重码累积成功概率、短重码维数有关,并提出采用Hilbert扫描序列以提高重码累积成功概率比率来提升检测性能的盲隐写检测方法。该方法在图像LSB Hilbert扫描序列中,基于码元相同短重码统计量的分布特征,通过Poisson分布显著性检验检测隐写信息,可充分利用Hilbert曲线良好的局部相关保持特性,不仅利用了相邻码元的相关性,还利用了局部区域码元的相关性。理论分析和实验结果表明了本文方法在有效控制虚警率的前提下,具有较好的隐写检测性能。By analyzing and proving the correlation between the detection capability of a short duplicate code statistical feature and the probability of cumulating short duplicate codes, the dimension of short duplicate codes, a method to im- proving the detection capability of a short duplicate code statistical feature was found. Then, a blind detection method for image steganography using short duplicate codes statistical model for Hilbert scanning sequences was proposed. The proposed method used Poisson distribution test to detect the stego-message based on the statistical feature of short dupli- cate code with same elements in LSB Hilbert scanning sequences. So, the proposed method could make full use of Hilbert curve to maintain the good properties of local correlation, and could not only use the correlation of adjacent elements, but also use the correlation of elements in local regions. Theoretical analysis and experiments show that the proposed method can effectively improve the detection rate under the condition of effectively controlling the false alarm rate.

关 键 词：盲隐写检测 码元相同短重码统计 Hilbert曲线扫描 相邻码元相关性 局部区域码元相关性 

分 类 号：TP911.73[自动化与计算机技术]