论文部分内容阅读
针对一种抗JPEG编码的密写方法,可由图象分块特性定义一个块效应参数作为统计检验量,并设定阈值从而判断秘密信息是否存在。由于基于块运算的抗JPEG密写和JPEG压缩编码都有块效应,为了完善基于块的密写分析法,提出了进一步的检测技术, 可有效地区分密写嵌入算法和JPEG压缩。该方法首先求图象的分块 DCT,计算不同频率位置上 DCT系数的直方图,并对直方图做DFT。JPEG压缩会造成非直流DCT系数直方图的离散性和准周期性,造成直方图“频谱”的高频分量提升并呈明显的等间距尖峰状;另一方面,通过改变图象块均值嵌入秘密信息的密写方法并不明显地改变非直流分量的直方图特性。基于这一原理,可通过检测其直方图频谱来区分JPEG和密写,从而有效地避免了密写分析中因图象曾被 JPEG压缩而引起的虚警。实验表明这一方法简单有效。
For a hard-coded anti-JPEG secret method, a block effect parameter can be defined as the statistical test quantity by the image block feature, and the threshold can be set to judge whether the secret information exists or not. Since block-based anti-JPEG JPEG and JPEG compression have block effects, in order to improve the block-based secret-to-close analysis method, a further detection technique is proposed, which can effectively distinguish between encryption algorithm and JPEG compression. In this method, the block DCT of the image is first calculated, the histogram of the DCT coefficients at different frequency positions is calculated, and the histogram is made DFT. JPEG compression causes discrete and quasi-periodic histograms of non-DC DC coefficients, resulting in a high-frequency component of the “spectrum” of the histogram that is promoted and appears sharply spiked at equal intervals; on the other hand, the embedding of the secret by changing the mean of the image blocks The method of hard-copy information does not significantly change the histogram characteristics of non-DC components. Based on this principle, the histogram spectrum can be detected to distinguish between JPEG and secret key, which effectively avoids the false alarm caused by JPEG compression in the close-write analysis. Experiments show that this method is simple and effective.