Optimization of filtered back-projection for a Rayleigh task

We consider the performance of the Hotelling Observer (HO) on a Rayleigh task in order to characterize the resolution properties of alternative implementations of the filtered back-projection (FBP) algorithm. Two alternative strategies, Hanning filtering in the reconstruction and Gaussian smoothing post-reconstruction, are considered. We find that use of the HO provides insight into the properties of these alternatives that would be lost through subjective image quality evaluation.