Minimization of fixed pattern noise in photon event counting imaging

Abstract

Low light-level ultraviolet and optical imaging with a photon counting image intensifier coupled to a charge coupled device camera generally results in varying levels of fixed pattern noise in the image. Here, we demonstrate that this can be minimized by the appropriate choice of photon event centroiding algorithm. We compare the fixed pattern noise generated by a center of gravity centroiding algorithm, a Gaussian centroiding algorithm, and a hybrid centroiding algorithm which uses center of gravity centroiding when one wing is zero, and Gaussian centroiding otherwise. This approach yields the best image quality with a lower fixed pattern noise parameter (9.99%) than the sole use of Gaussian centroiding (16.4%), and there is no need for a look-up table correction. In addition, the hybrid algorithm also yields maximum detective quantum efficiency by overcoming small pulse centroiding failure associated with Gaussian centroiding. The digitization error when recording the events is modeled with a Monte Carlo simulation and discussed. It is found that a center of gravity algorithm produces not only significant fixed pattern noise, but also pulse height dependent x̄ positions. For a Gaussian centroiding algorithm the x̄ positions are independent of the pulse height, the fixed pattern noise is low and the digitization error only yields a small increase of the fixed pattern noise parameter. This shows that while there is a limit to centroiding accuracy due to the digitization error, the appropriate choice of centroiding algorithm is a much more important factor to minimize fixed pattern noise.