量化不一致性的JPEG合成图像盲检测

Blind Detection of the JPEG Composite Image Based on the Inconsistencies of Quantization

通过估计量化因子获取合成图像初次压缩因子和第三次量化后的图像区域的相似度,分别利用块效应映射函数、灰度转换函数可视化相似度和矫正输出图像,最后利用立方回旋插值法使块映射图与篡改图像尺寸相同.以JPEG合成图像为例的实验表明:该算法只需寻找出量化差异即能完成对篡改区域的定位。当二次压缩因子与初次压缩因子之差为5时,能取得比较明显的检测效果.

By estimating quantification factors of the image,this paper obained the first image compression factor and similarity of image regions after third quantification.Visualization of similarity measure and correcting output image were performed by block mapping functions and gray transformation functions,respectively.Finally,using the cubic swing interpolation method made the size of piece of mapping same with the doctored part of the image.The experiment showed that the algorithm could position the doctored area with the needs to find out the quantitative differences.When the difference of secondary compression factor and the initial one was 5,obvious detection effect could be obtained.