Enhancing Source Camera Identification Based on Multiplicative Denoising Filter
Qiang Rao, JianJun Wang, Liming Zhang · 2016
Photo-response non-uniformity (PRNU) noise has been recognized as a unique fingerprint of digital cameras. However, how to extract precise PRNU noise from a given image is a main challenge. Previous literatures assume PRNU noise is a white Gaussian noise (WGN) and extracted by an additive denoising filter (ADF). However, experiments have demonstrated that the PRNU noise is a multiplicative noise. In this paper, an approach is proposed to extract more accurate PRNU noise through multiplicative denoising filter (MDF). Specifically, PRNU noise is modeled as a WGN multiplied by image content and extracted by proposed MDF based on linear minimum mean squared error (LMMSE) estimation. Experimental results demonstrate the proposed method outperforms, or at least performs comparably to, the state-of-the-art methods.