Robust image source camera identification using Bayesian MRF and MAP with Wiener soft thresholding (BMRFW)
Pabitra Roy, Shyamali Mitra, Abdur Rahaman Sardar, Nibaran Das · The Imaging Science Journal · 2025
In this study, we have proposed novel source camera identification (SCI) method that efficiently handles the challenges of reducing scene detail distortion. This distortion highly affects the Sensor Pattern Noise (SPN) extraction and reduces the recognition rate, particularly with smaller images. To address these issues, we have developed a new denoising filter that uses Bayesian Markov random field (MRF), maximum a posteriori (MAP) estimation, and Wiener soft thresholding for efficient extraction of pure SPN by suppressing stronger and less reliable components. Existing methods often fail to balance between accuracy and robustness, whereas our method not only outperforms existing techniques but also excels in handling images shared on platforms like Facebook and WhatsApp, offering a reliable and robust solution for real-world source camera identification.