Effective Source Camera Identification based on MSEPLL Denoising Applied to Small Image Patches
Wen-Na Zhang, Yunxia Liu, Zeyu Zou, Yun-Li Zang, Yang Yang, Ngai-Fong Law · 2019
Sensor Pattern Noise (SPN) has proven to be an effective fingerprint for source camera identification, while its estimation accuracy heavily relies on denoising algorithm. In this paper, an effective source camera identification scheme based on Multi-Scale Expected Patch Log Likelihood (MSEPLL) denoising algorithm is proposed, firstly. With enhanced prior modeling across multiple scales, MSEPLL can accurately restore the original image. As a consequence, estimated SPN is less influenced by image content. Secondly, the source camera identification problem is formulated by hypothesis testing, where normalized correlation coefficient is adopted for SPN detection. Finally, the effectiveness of the proposed method is verified by abundant experiments in terms of identification accuracy as well as receiver operating characteristic. Performance improvement is more prominent for small image patches, which is more conducive to real forensics applications.