摘要
在对现有盲检测算法分析研究改进的基础上,提出一种新的JPEG图像信息隐藏盲检测方法.该方法把JPEG图像的DCT系数分解到小波域,计算出各个子带小波系数的一系列高阶统计特征,组成72维特征向量,引入特征缩放技术将特征向量进行缩放,并用支持向量机进行分类.针对CorelDraw图像库4种常用的隐写方法进行实验,结果表明该方法在4种不同隐写方式下的平均检测正确率均达到91%(嵌入率为100%)以上.
On the basis of analysis, investigation, and modification of existing blind detection algorithm, a new method for blind detection of secret information in JPEG image was proposed. This method was used to decompose the DCT coefficients of JPEG images in wavelet domain and then calculate a series of higherorder statistical features of the sub-band wavelet coefficients, making up 72-dimensional featui'e vector. By introducing feature zooming technique, this 72-dimensional feature vector was scaled up and down, and classified with support vector machine. Experimental results of the images taken from CorelDraw image database showed that the mean correct detection rate was greater than 91% in four kinds of different steganography mode (embedding rate being 100%).
出处
《兰州理工大学学报》
CAS
北大核心
2009年第5期85-88,共4页
Journal of Lanzhou University of Technology
基金
甘肃省自然科学基金(0803RJZ024)
关键词
隐写分析
小波变换
高阶统计量
特征缩放
支持向量机
steganalysis
wavelet transform
higher-order statistics
feature scaling
support vector machine