一种基于CFA插值系数熵的图像来源取证算法

Blind CFA Interpolation Detection Based on The Entropy

研究分析了数码相机内部的CFA插值算法,利用数字图像本身对数码相机的CFA插值系数进行估计,同时将熵理论引入其中,以插值系数熵作为特征值,使用BP神经网络作为分类器,提出了一种基于CFA插值系数熵的图像来源取证算法。实验表明:该算法能够较准确地对不同相机来源的数字图像进行鉴别,同时对JPEG重压缩图像的检测率也有所提高。

The study analyzed the digital camera＇s internal CFA interpolation algorithm, estimated the CFA interpolation coefficients of the digital camera by digital image itself. While the entropy theory was introduced to this paper, used entropy of interpolation coefficient as a feature and the BP neural network as a classifier, and then proposed an image source forensics algorithm based on the entropy of CFA interpolation coefficients. Experiments showed that the algorithm could identify images from different cameras more accurately, and had increased the detection rate to the re-compressed JPEG images.