A maximum likelihood expectation maximization iterative image reconstruction technique for mask/anti-mask coded aperture data

We present a method to use mask/anti-mask coded aperture data with maximum likelihood expectation maximization (MLEM) image reconstruction. The mask/anti-mask approach eliminates “unmodulated” data, improving image quality when backgrounds, room scatter, or noisy detectors are significant. MLEM permits complex detector response models, desirable in gamma-ray or fast neutron imaging with thick masks, near-field imaging, or tomographic reconstruction. Subtracted mask/anti-mask data is not Poisson distributed, and cannot be used with MLEM. Instead, we treat unmodulated data as generated by source terms indexed by detector pixel, so that MLEM converges to simultaneous estimates of the true image and the unmodulated event rates.