Robust PRNU estimation from probabilistic raw measurements

Images captured by digital cameras undergo various in-camera processing such as JPEG compression, white balancing, power transforms and other operations to map raw data into nonlinear small gamut image. Due to nonlinear transformation, artifacts or signatures used for camera identification also undergo a significant change. Photo Response Non-Uniformity (PRNU), proved to be useful for uniquely identifying the camera, also undergoes same in-camera operations. Hence estimation of PRNU is affected which often leads to rise in false identification. In this work, we develop a novel algorithm for robust estimation of PRNU from probabilistically obtained raw data. Since not all cameras provide raw data as their output, we compute raw data from the JPEG output using probabilistic color de-rendering procedure. The estimated raw data is modeled as a Poisson process, and Maximum Likelihood Estimation (MLE) is used for PRNU estimation. We then use our estimate of PRNU for identifying the camera using images. We also compare the performance of our algorithm with other state-of-the-art algorithms. Additionally, we demonstrate the robustness of estimate obtained by localizing the forgery in images. The extensive experimental analysis is performed over thousands of patches from various cameras to illustrate the efficiency of proposed approach, which effectively overcomes the state-of-the-art.