基于首数字定律的图像高斯模糊篡改检测算法

IMAGE GAUSSIAN BLUR TAMPERING DETECTION ALGORITHM BASED ON LEADING DIGIT LAW

针对现有图像篡改操作中常见的高斯模糊篡改检测算法普适性不强、效率不高的缺点,提出一种基于首数字定律的检测算法。首先,提取图像的RGB三个彩色通道的离散余弦变换(DCT)域交流分量(AC系数)和梯度的首位有效数字统计特征。然后,利用支持向量机分类器训练并进行分类。最后,用从标准数据库下载大量图像与自拍图像对本算法进行验证。实验表明,该算法可以有效检测出自然图像是否经过高斯模糊篡改。

In this paper,we propose a new detection algorithm,it is based on leading digit law,to cope with the defects of poor universality and inefficiency of Gaussian blur tampering detection algorithm which are common in existing image tampering operations. First,it extracts the statistics features of valid leading digits in regard to the AC coefficients of DCT-domains and the gradients from RGB three channels of the image respectively. Secondly,it employs the support vector machines（ SVM） classifier to train and to classify those unknown examples. Lastly,it uses a number of images downloaded from standard database and the self-shot photos to verify this algorithm. Experiments show that the proposed algorithm can effectively detect whether a natural image has been tampered by Gaussian blurs operation.