Texture-aware sensor pattern noise estimation for accurate camera source identification

This paper presents a novel camera source identification method, termed Improved Camera Source Identification using Wavelet Noise Residuals and Texture Filtering (ICSI-WNRTF) . The proposed approach addresses a key limitation in traditional Sensor Pattern Noise (SPN)-based techniques by excluding high-textured regions that introduce scene-dependent noise and reduce identification accuracy. By leveraging wavelet-based noise residuals and a texture-aware filtering mechanism, the method enhances the robustness and discriminative power of the extracted camera signature. The ICSI-WNRTF method was evaluated using the VISION dataset under both model-wide and camera-specific scenarios. Experimental results demonstrate that the proposed method significantly outperforms the baseline TCSI-WNR and several state-of-the-art techniques in terms of accuracy, precision, recall, and false positive/negative rates. Specifically, the method achieved 99% accuracy in model-wide evaluation, with 2.7% FNR and 0.48% FPR, and 98% accuracy in camera-specific analysis, with 12.6% FNR and 0.85% FPR. A hybrid thresholding strategy was also introduced to improve classification reliability. These findings confirm that excluding high-textured regions from SPN estimation leads to more accurate and robust camera source identification, offering a promising solution for digital image forensics and multimedia authentication applications.