CAJ | 학술논문

通过大量实验发现,基于块中心跳转的二值文本图像密写算法,其二值文本干净图像和密写图像的游程统计特性差别较大,而二值文本密写图像和二次密写图像的游程统计特性差别较小.基于该规律提出一种二值文本图像信息隐藏分析方法,对待测图像进行二次密写,计算待测图像和密写图像游程统计函数的差异值,当差异值大于门限时,待测图像是干净图像,否则是密写图像.通过分类实验的方法确定了几种典型字体的判决门限.实验结果表明,利用该方法和相关判决门限,密写检测率可达99.9%.
통과대량실험발현,기우괴중심도전적이치문본도상밀사산법,기이치문본간정도상화밀사도상적유정통계특성차별교대,이이치문본밀사도상화이차밀사도상적유정통계특성차별교소.기우해규률제출일충이치문본도상신식은장분석방법,대대측도상진행이차밀사,계산대측도상화밀사도상유정통계함수적차이치,당차이치대우문한시,대측도상시간정도상,부칙시밀사도상.통과분류실험적방법학정료궤충전형자체적판결문한.실험결과표명,이용해방법화상관판결문한,밀사검측솔가체99.9%.
Through massive experiments, a law is fownd. It is for the binary text image embedding algorithm which hiding information by flipping center pixel of the image block, the statistics of run length between cover image and the image "after test embedding is difference obviously, but which between stego image and the image after test embedding is similar. Based on this law, a novel technique for the steganalysis of binary text image is presented. In which a test message is embedded in the text image, the run length statistic difference between text image and the image after test embedding is calculated. When the difference exceeds the threshold value, the image is cover image, otherwise it's stego image. Through the classification experiment, the threshold of several typical characters is identified. Experimental results show that detection rate can reach 99.9% by using of the proposed method and the related threshold.