双JPEG压缩图像中初始量化系数的估算方法

Estimation Method of Primary Quantization Coefficient in Double Compressed JPEG Images

为有效分析双JPEG压缩的统计特性,论述了双JPEG压缩现象,并对某一特定位置的DCT系数的直方图进行了分析,从中找出某些特性,如:双峰,零值点等.针对不同的双压组合,其DCT直方图的特性会有所不同,利用支持向量机(SVM)对其进行分类,从分类的结果来看,分类的正确率普遍能够达到98%以上,再用训练好的模式去估计双JPEG压缩图像中初始量化系数.

This paper discusses the phenomenon of double-JPEG compression with a view to effectively analyzing its statistical properties. Based on the analysis of specific location of the DCT coefficient of histogram, some characteristics such as double peaks and zero points were obtained. Double-JPEG compression was classified according to diverse characteristics of its DCT histogram with the use of support vector machines （ SVM）. The result indicates that more than 98 percent of the classification is correct. Then, the well-trained pattern was employed for the purpose of estimating primary quantization coefficient in double compressed JPEG images.