Abstract

Information describing a scene must often be inferred from its spatial autocorrelation function (or its power spectrum). Such information can be used to establish constraints or trial solutions in iterative image reconstruction, or it can be used to fit scene object models directly to nonimage data. The most basic scene information consists of the number of objects present and their relative positions, which is prerequisite to describing the individual objects. Unfolding point-object positions from a noise-free autocorrelation has been addressed previously. Many applications involve extended objects, however, inviting the problematic overlap of autocorrelation features. Furthermore, data are rarely noise-free. A method of logically unfolding object positions from an autocorrelation is described, and the impact of noise, redundancy, and finite object size on the process is assessed.

© 1993 Optical Society of America

Extracting properties of multiple-scene objects from autocorrelation features

George E. Cipperly
Appl. Opt. 32(29) 5775-5781 (1993)

Imaging a randomly moving object from quantum-limited data: applications to image recovery from second- and third-order autocorrelations

Timothy J. Schulz and Donald L. Snyder
J. Opt. Soc. Am. A 8(5) 801-807 (1991)

Extracting input-line position from Hough transform data

Jeffery Richards and David P. Casasent
Appl. Opt. 30(20) 2899-2905 (1991)

Reconstruction of refractive index fields by using relative fringe data

Yan Zhang and Gary A. Ruff
Appl. Opt. 32(16) 2921-2926 (1993)

Multiple-object detection with a chirp-encoded joint transform correlator

Qing Tang and Bahram Javidi
Appl. Opt. 32(26) 5079-5088 (1993)