Modeling and Real-Time Estimation of Signal-Dependent Noise in Quantum-Limited Imaging

Many established and emerging image processing applications rely on quantum-limited imaging, i.e., imaging in extremely poor illumination. At this, images are corrupted by severe signal-dependent Poisson noise. For optimal noise reduction the noise characteristics must be estimated and integrated into the method. Common noise estimators, however, assume Gaussian noise which is not signal-dependent. In this paper, we describe the modeling process exemplarily for low-dose medical X-ray imaging. In this context, we formulate functional models for detector images and images which have undergone nonlinear white compression prior to further processing. Furthermore, we present a robust estimator for signal-dependent noise suited for real-time applications. Key–Words: Quantum noise, noise estimation, X-ray imaging, white compression