基于人工神经网络的数字图像盲取证

Blind Detection Algorithms for Forged Images Based on Artificial Neural Networks

数字图像盲取证技术是一项通过图像本身统计特性来鉴别图像是否被篡改的技术。文章基于数码相机的成像原理,利用人工神经网络来预测成像过程中由于CFA(color filter arrays)插值算法导致图像像素点之间出现的相关性关系,由于没有篡改的图像采用的CFA插值算法是相同的,而被篡改的图像则可能存在两种以上的CFA插值算法,两种情况下图像预测误差会出现差异,本算法利用这种差异实现数字图像的盲取证。实验结果表明本算法具有一定的有效性。

Blind detection algorithms for images is a technique which uses image itself statistical features to identify the forged images.In this paper the authors bring forward a new algorithms which is based on imaging principle of digital camera,using artificial neural networks to forecasting the relationship between the relevant pixel.The image which has not been forged has only one interpolated algorithms,but the forged one may have two or more,so if using artificial neural networks to forecast two types of these images we will get different results,by this difference can find forged image.The result of simulation implies that this calculation has some effects.