Abstract: According to the analysis on the effect of the texture complexity on steganalysis, the authors proposed a data hiding algorithm based on texture complexity and pixel-value difference. Firstly, the image was decomposed into blocks. Then, the texture complexity of each block was calculated and blocks with high complexity were selected. Then the pixel-value difference on two directions of the selected blocks was computed to determine the edge areas. The secret messages were only embedded in the pixel-value difference above the threshold. When compared with several steganographic methods on different image databases, the experimental results show that the proposed method outperforms previous steganographic methods such as LSB matching and adaptive data hiding in edge areas of images on the capability of resisting universal blind detection, such as wavelet high-order statistics analysis.

Key words: steganography, universal blind detection, texture complexity, edge area, LSB matching

摘要： 通过分析纹理复杂度对图像隐写分析的影响,提出一种基于纹理复杂度和像素差分的隐写算法。首先,将图像划分为大小相同的分块,通过分块中起点位置的不同计算分块中不同小块的纹理复杂度,选择复杂度最高的图像小块;然后,计算小块的相邻像素差分值,并根据秘密消息的长度和不同方向上的差分值分布确定嵌入阈值,在大于该阈值的差分值中嵌入秘密消息。对多种嵌入算法在不同图像库上做通用盲检测实验,实验结果表明,该算法在抵抗小波高阶统计量分析等盲检测算法能力方面,优于LSB匹配、图像边缘自适应嵌入等隐写算法。

关键词: 数字隐写, 通用盲检测, 纹理复杂度, 边缘区域, LSB匹配