Adaptive photo-response non-uniformity noise removal against image source attribution

Abstract

The main objective of image source anonymization is to protect the identity of the photographer against any attempts to identify the source camera device through PRNU noise analysis. One way of impeding image source attribution is to suppress the PRNU noise as much as possible. In this paper, we introduce an improvement on the existing adaptive photo-response non-uniformity (PRNU) denoising method against source camera identification. We evaluate the performance of the proposed method with substantial experimental analysis. We also provide anonymization benchmarks with other source anonymization techniques. The benchmarks' results show that the proposed method outperforms the adaptive PRNU denoising methods for various cameras including compact and smartphone in terms of speed and image quality. The experimental analysis also shows that it is possible to impede source camera identification by PRNU noise suppression even under extreme attack conditions.