Abstract: To improve the security of Bit-Plane Complexity Segmentation (BPCS) steganography algorithm, the authors proposed a BPCS steganography algorithm against statistical analysis. Firstly, an image was divided into blocks to calculate the information entropy and the wavelet contrast. Then the blocks were classified by Fuzzy C-Means (FCM) and each bit-plane block was set different complexity thresholds based on the classification result and a random number. Finally, if the similarity of secret data and carrier bit-plane block was less than 0.5, the secret data bit-plane block was negated to replace the carrier bit-plane block. The experimental results show that the proposed algorithm can effectively resist the detection of statistical analysis of the complexity histogram. Furthermore, the visual imperceptibility of stego image has been greatly improved.

Key words: steganography, Bit-Plane Complexity Segmentation (BPCS), Fuzzy C-Means (FCM), complexity histogram, wavelet contrast

摘要： 为提高位平面复杂度分割(BPCS)密写算法的安全性，提出了一种抗统计分析的BPCS密写算法。首先，将载体图像分块，计算图像块的信息熵和小波对比度；然后，利用模糊C-均值(FCM)对其进行分类,根据分类结果及随机数确定每个位面小块的复杂度阈值；最后，将载体和秘密数据位面小块相似程度小于0.5的秘密数据位面小块取反，再替换载体位面小块。实验结果表明，该方法能有效地抵抗复杂度直方图统计分析的检测，同时载密图像视觉不可感知性有了很大的提高。

关键词: 密写, 位平面复杂度分割, 模糊C-均值, 复杂度直方图, 小波对比度