Enhancing source camera identification using weighted nuclear norm minimization de-noising filter

Abstract

Photo-response non-uniformity noise (PRNU) is widely accepted as fingerprint (FP) of digital camera. However, extraction of PRNU from given images is still a challenging task. In the previous literature, number of de-noising filters has been used for PRNU extraction. However, it is observed that PRNU extracted by existing de-noising filters contains high-frequency (edges and texture) details of the image. This increases false rejection rate in source camera identification (SCI) process. In this work, we have used weighted nuclear norm minimization (WNNM)-based de-noising filter for PRNU extraction. The PRNU extracted by WNNM-based de-noising filter contains least amount of scene details. Experimental results demonstrate the proposed method outperforms, or at least performs comparably to, the state-of-the-art methods.