Abstract: Aiming at the problem that existing steganalysis algorithm has poor detection performance, this paper presents a universal steganalysis algorithm based on multi-wavelet statistics feature. Image is decomposed to multi-scale and multi-direction through multi-wavelet transform. Then two kinds of statistics features are got through every sub-image. The Support Vector Machine(SVM) is trained on these statistic features to construct a universal steganalysis. Results show that this algorithm can catch more effective features of image and it has correct detection rate compared with Farid algorithm.

Key words: steganalysis, multi-wavelet, Support Vector Machine(SVM)

摘要： 针对现有隐写分析算法检测性能较差的问题，提出一种基于多小波统计特征的通用隐写分析算法。该算法采用多小波变换对样本图像进行多尺度分解，在各子带中提取广义高斯模型和多小波高阶统计特征，通过结合支持向量机分类器对大量图像样本进行隐写分析。结果表明，与经典的Farid算法相比，该算法提取的多小波统计特征更有效，且具有更高的检测率。

关键词: 隐写分析, 多小波, 支持向量机

CLC Number: 

• TP391