Abstract
An imaging photon-counting detector is used to achieve automatic sorting of two image classes. The classification decision is formed on the basis of the cross correlation between a photon-limited input image and a reference function stored in computer memory. Expressions for the statistical parameters of the low-light-level correlation signal are given and are verified experimentally. To obtain a correlation-based system for two-class sorting, it is necessary to construct a reference function that produces useful information for class discrimination. An expression for such a reference function is derived using maximum-likelihood decision theory. Theoretically predicted results are used to compare on the basis of performance the maximum-likelihood reference function with Fukunaga–Koontz basis vectors and average filters. For each method, good class discrimination is found to result in milliseconds from a sparse sampling of the input image.
© 1986 Optical Society of America
Full Article | PDF ArticleMore Like This
Thomas A. Isberg and G. Michael Morris
J. Opt. Soc. Am. A 3(7) 954-963 (1986)
G. Michael Morris
Appl. Opt. 23(18) 3152-3159 (1984)
J. Anthony Tyson
J. Opt. Soc. Am. A 3(12) 2131-2138 (1986)