Abstract: Doubly compression introduced by JPEG image forgery usually causes the original compression properties of the tampering region change. Therefore, it can utilize the compression property inconsistency to detect image forgery. Using the principle, this paper proposes a JPEG image forgery detection algorithm based on quantization noise. The paper blocks the image and calculates the quantization noise of each block. It gets the probabilities of the quantization noise of each image block satisfying uniform distribution and Gaussian distribution. As a consequence, the tampered regions doubly compressed are detected. The experimental results indicate that this algorithm can detect the tampering of JPEG images with double compression and locate the tampering region.

Key words: JPEG image, quantization noise, forgery detection, Gaussian distribution

摘要： JPEG图像篡改引入的双重压缩会导致篡改区域的原始压缩特性发生改变，因此可以利用篡改区域压缩特性的不一致性来检测图像的篡改。利用该原理，提出了一种基于量化噪声的JPEG图像篡改检测算法。算法对待检测图像进行分块，计算每块的量化噪声，求取图像块的量化噪声服从均匀分布和高斯分布的概率，从而检测出篡改过的双重压缩区域。实验结果表明：该算法能有效检测双重压缩的JPEG图像篡改，并能定位出篡改区域。

关键词: JPEG图像, 量化噪声, 篡改检测, 高斯分布