We can greatly reduce image artifacts in our pixellated CdZnTe arrays by ma
pping imperfect regions with a narrow collimated beam of gamma rays. Portio
ns of our detectors produce signals that agree well with simulations of gam
ma-ray interactions, but there are many examples of structures in the mater
ial that respond unpredictably to gamma rays. We mapped some of these imper
fect regions using 60 and 140 keV gamma-ray beams, recording a 7 x 7 set of
pixel signals for each interaction. The pixel pitch was 380 mu m. We used
the mapped data to estimate the probability density function (PDF) of the p
ixel signals for each interaction position. Images were taken on the mapped
sections, storing each gamma; ray as a list of pixel signals. Images could
be formed by either estimating each gamma-ray interaction position individ
ually or using the entire set of image data in a single iterative computati
on using the expectation-maximization (EM) algorithm. At 60 keV individual
interaction positions were estimated by fitting the data to a Gaussian PDF,
correcting the artifacts and giving sub-pixel resolution of less than 150
mu m in some regions. At 140 keV applying the EM algorithm was necessary fo
r improving the images. (C) 1999 Published by Elsevier Science B.V. All rig
hts reserved.