Camera brand and model identification using moments of 1-D and 2-D characteristic functions

Camera brand and model identification has become one important task of image forensics. Most of the research on this topic focuses on only one or two parts of camera inner structure. In this paper, we propose a universal image statistical model which takes the whole image formation pipeline of cameras into consideration. By examining their comprehensive effects on the formulated images, our assumption is that any difference of the parts of the image formation pipeline can result in the statistical difference of the output image. Moments of 1-D characteristic functions generated from the given image, its JPEG 2-D array, their prediction-error 2-D arrays, and all of their three-level wavelet subbands, and moments of 2-D characteristic functions generated only from JPEG 2-D array accordingly are used to build the statistical model for classification. Our experimental works have verified the effectiveness of this proposed method.