The problem posed in this paper is that of restoring a Poisson-point-process intensity that has been degraded by a band-limiting filter followed by a truncation of the signal. The approach is to derive a maximum-likelihood estimate from the count data of the degraded point process. The expectation-maximization algorithm is used to realize this estimate, while the derivation of this algorithm is an extension to previous developments by Shepp and Vardi [
IEEE Trans. Med. Imaging MI-2,
1982)], Snyder et al. [
IEEE Trans. Nucl. Sci. NS-28,
1981)], and others used for positron-emission tomography. We also extend our own work reported earlier by considering the truncated signal, which is analogous to practical situations in both two- and three-dimensional microscopy in which the image of the specimen has been truncated. Computer simulations with one-dimensional and two-dimensional signals demonstrate such a reconstruction with reasonable success. The plausibility of doing such a reconstruction is explained in that for the noiseless case the transformation characterizing the degradation is invertible.
© 1989 Optical Society of America
Equations on this page are rendered with MathJax. Learn more.