在分析通用盲检测方法原理的基础上,提出一种通过修改整数小波高频系数实现消息嵌入的隐写法.该方法使用小波去噪门限作为消息嵌入的阈值,在大于该门限的小波高频系数中嵌入消息,并在阈值处引入直方图补偿策略使小波系数直方图近似保持不变.由于通用盲检测方法大多利用概率密度函数的变化实现检测,因此该文提出的方法可以获得对通用盲检测算法的抵抗力.实验结果表明,该方法在抵抗小波高阶统计量分析、直方图特征函数质心等盲检测法的能力方面,优于LSB匹配、像素值差分等隐写法.
Based on the principle of blind detection techniques,a data hiding scheme that modifies high frequency integer-wavelet coefficients is proposed.Coefficients of a subband above a wavelet denoising threshold are chosen for data embedding.To preserve the histogram of wavelet coefficients,an adjustment strategy is introduced around the threshold.Since most blind detection methods select classification features according to differences of statistical distributions between the cover and stego images,the proposed method can resist blind detection.Experimental results show that the method outperforms previous steganographic methods such as LSB matching and pixel-value differencing in terms of the capability of resisting current typical universal blind detecting methods.