Camera model identification based on the generalized noise model in natural images

The goal of this paper is to design a statistical test for the camera model identification problem. The approach is based on the generalized noise model that is developed by following the image processing pipeline of the digital camera. More specifically, this model is given by starting from the heteroscedastic noise model that describes the linear relation between the expectation and variance of a RAW pixel and taking into account the non-linear effect of gamma correction. The generalized noise model characterizes more accurately a natural image in TIFF or JPEG format. The present paper is similar to our previous work that was proposed for camera model identification from RAW images based on the heteroscedastic noise model. The parameters that are specified in the generalized noise model are used as camera fingerprint to identify camera models. The camera model identification problem is cast in the framework of hypothesis testing theory. In an ideal context where all model parameters are perfectly known, the Likelihood Ratio Test is presented and its statistical performances are theoretically established. In practice when the model parameters are unknown, two Generalized Likelihood Ratio Tests are designed to deal with this difficulty such that they can meet a prescribed false alarm probability while ensuring a high detection performance. Numerical results on simulated images and real natural JPEG images highlight the relevance of the proposed approach.