噪声特征与EXIF信息相关性的图像篡改鉴定

Detecting Forgeriesby Correlation between Image Noise Features and EXIF Parameters

建立了照片可交换图像文件(EXIF)信息中的5个常用参数与图像统计噪声之间的相关性,利用这种相关性对图像的真实性做出判断.通过提取训练图像的统计噪声特征及5个EXIF特征,建立了二者的相关性,利用此相关性估计待检测图像的EXIF信息,再与照片本身的EXIF信息进行对比,进而判断其是否遭到篡改.实验结果显示,利用提出的方法对图像是否经过照片滤镜和模糊等修改进行检测,得到了较好的检测结果.

In correlation between five common exchangeable image file （EXIF） parameters and image statistical noise for digital photos is built. The authenticity of the image is judged by this correlation. The statistical noise and five EXIF features from the training image are extracted, following the correlation es- tablished. Make use of the correlation to estimate EXIF information of testing images which will be com- pared with the real values. The comparison will determine the authenticity of testing images. Experiments on both tampered images and real images indicate the effectiveness of the proposed method.