基于JPEG量化失真的合成图像盲检测

Blind detection of synthetic image based on JPEG quantization distortion

主要针对JPEG图像合成伪造,提出了一种基于量化失真的合成图像盲检测算法。首先针对合成图像以JPEG和非JPEG不同的存储方式,分别估计原始量化矩阵;然后用估计的原始量化矩阵对合成图像再压缩,计算压缩前后的量化失真;最后通过判断合成图像不同区域量化失真的大小,实现窜改区域的自动检测和定位。实验结果表明:算法能有效地检测JPEG和非JPEG两种不同存储方式的合成图像。

This paper proposed an algorithm to blindly detect synthetic image based on JPEG quantization distortion.Firstly,it estimated the original quantization matrix of the synthetic image in terms of its storage format,JPEG or non-JPEG.Secondly,it recompressed the synthetic image using the estimated original quantization matrix,calculated the quantization distortion of the synthetic image before and after compression.Finally,it judged the size of quantization distortion for different region of synthetic image,to realize automatic detection and location of its tampered regions.Experimental results show that the proposed algorithm is effective in detecting synthetic image stored in both JPEG and non-JPEG.