A simple algorithm for restoring images from pixellated radiation detectors

Pixellated radiation detectors suffer from a variety of non-ideal characteristics, many of which vary significantly among pixels. These include variations in gain, different background or dark-signal levels, and variations in noise characteristics. A simple method that corrects for these effects in a unified manner is described in this work. The method seeks a non-negative solution that minimizes a penalized, weighted least squares objective function. The penalty function encourages smoothness in the estimates and the variable error weighting allows measurements having disparate uncertainties to be appropriately combined. Solution is via an iterative Gauss-Seidel method with convergence usually occurring in a few iterations.