利用JPEG压缩特性的合成图像检测

提出一种利用压缩特性进行JPEG合成图像检测的方案。由多幅JPEG图像内容拼接而成的合成图像中往往存在质量因子不一致性或分块位置不一致性，这两种不一致性可以作为检测JPEG合成图像的有力证据。该文所提方案先对待检测图像进行尝试性的再压缩并考察失真程度，以估计主要内容曾经历过的压缩情况，然后根据再压缩在不同区域引起的失真程度的起伏变化识别合成图像。该方法还可区分合成图像内容的不同来源，并分析质量因子不一致性或分块位置不一致性，为揭示图像伪造行为提供辅助信息。

Detection of Composite Images Based on JPEG Compression Properties

This paper proposes an image forgery detection scheme based on JPEG compression properties, which aims to identify a special type of images that are composed of contents from different JPEG images. Inconsistency in compression qualities and block boundaries in a suspicious image can provide a clue for forgery detection. By recompressing the image with different quality factors and analyzing the induced distortion, we can estimate the quality factor initially used in the main area of the image. Thus, inconsistency can be found from fluctuation of distortion energy in different areas. The proposed scheme is capable of identifying objects from different origins, and provides a useful means of digital forensics.