Michael S. F. Kirk (NASA Goddard Space Flight Center / Catholic University of America)
Poisson-Gaussian Noise Modeling in Solar Images
All digital images are corrupted by noise. In most solar imaging, we have the luxury of high photon counts and low background contamination, which when combined with carful calibration, minimize much of the impact noise has on the measurement. Outside high-intensity regions, such as in coronal holes, the noise component can become significant and complicate feature recognition and segmentation. We create a practical estimate of noise in the Solar Dynamics Observatory’s Atmospheric Imaging Assembly (SDO AIA) images across the detector CCD. A Poisson-Gaussian model of noise is well suited in the digital imaging environment due to the statistical distributions of photons and the characteristics of the CCD. Using the dark and flat field calibration images, the level-1 AIA images, and readout noise measurements, we construct a maximum-a-posteriori estimation of the expected error in the AIA images. These estimations of noise not only provide a clearer view of solar features in AIA, but they are also relevant to error characterizations of other solar images.
