基于块效应的JPEG合成图像盲检测

作为应用最为广泛的图像压缩格式,JPEG图像格式活跃于各个信息渠道之中,随之而来的针对JPEG图像的合成和修改也日益增多.由于JPEG图像中的块效应的存在,很多修改痕迹被掩盖,使得JPEG合成图像盲检测一直是修改图像检测的瓶颈之一.但是,在JPEG图像的合成过程中,为了达到较好的视觉效应,在修改区域往往会出现8×8子块的错位,这一错位在频域中表现为原有波峰的衰减或分裂.通过对这一特征的分析,提出了一个新的基于块效应的JPEG合成图像的盲检测算法,并且通过实验证实了其有效性.

Blind Detection of Merged JPEG Images Based on the Block Grids Mismatch

Blind JPEG forensic image is one of the bottlenecks in the detection of doctored images,since the blocking process covers most tampering trails.The 8×8 block grids of JPEG image combine during merge operation.To achieve a good vision effect,the grids are often mismatched.This causes some multi-peak phenomenon in frequency spectrum of block grids.Thus,a new blind detection algorithm of merged JPEG image is proposed based on this phenomenon.Experiment results demonstrate the availability of this algorithm.