On the influence of denoising in PRNU based forgery detection

To detect some image forgeries one can rely on the Photo-Response Non-Uniformity (PRNU), a deterministic pattern associated with each individual camera, which can be loosely modeled as low-intensity multiplicative noise. A very promising algorithm for PRNU-based forgery detection has been recently proposed by Chen et al. Image denoising is a key step of the algorithm, since it allows to single out and remove most of the signal components and reveal the PRNU pattern. In this work we analyze the influence of denoising on the overall performance of the method and show that the use of a suitable state-of-the art denoising technique improves performance appreciably w.r.t. the original algorithm.