摘要
QIM-JPEG2000作为以JPEG2000图像为载体的最新隐写术,采取设置恒域的方式,保持了小波系数的直方图,很好地抵御了基于直方图攻击的隐写分析方法.针对这种隐写术,提出了一种基于JPEG2000图像小波系数主分量分析的隐写分析方法.该算法在JPEG2000图像三级小波系数上进行主分量变换,分析主分量的统计特征在信息隐写前后的变化,提取小波系数主分量的均值、方差、斜度、峰值和熵组成特征向量,利用支持向量机进行载体图像和载密图像的分类,从而检测出JPEG2000图像是否嵌入了秘密信息.实验表明了本方法的可行性.
QIM-JPEG2000,the latest JPEG2000 steganography,had a way to set the constant domain to maintain the histogram of wavelet coefficients and resist the histogram-based steganalysis method.In response to this steganography,an information steganalysis methodology based on principal component analysis of DWT coefficients in JPEG2000 image was proposed.The algorithm transformed the principal component of 3rd DWT coefficients in JPEG2000 image,analyzed the difference between the statistical characteristics of the cover images and the stego images,extracted the mean,the variance,the slope,the peak,and the entropy of the principal components of wavelet coefficients as feature vectors.Then the algorithm used Support Vector Machines（SVM） to assort the cover images and the stego images.Thus the algorithm determined whether the JPEG2000 image embedded secret information or not.Experiments proved the feasibility of this method.
出处
《武汉大学学报（理学版）》
CAS
CSCD
北大核心
2010年第2期245-248,共4页
Journal of Wuhan University:Natural Science Edition
