Optical Spatial Filter Development.

Abstract

Abstract : The objective of this program was the design of optical spatial filters. Among the topics discussed in this report are: the proper role of correlation in target recognition and discrimination; noise considerations for the objective function used in designing spatial filters; correct computer modeling of the correlation process; a completely new approach to the design of general spatial filters as a constrained optimization question in all free parameters (all amplitudes and phases); the design of spatial filters whose entries are either zero or a continuous phase as an optimization question; the optimal discretization of spatial filters whose entries are either zero or a continuous phase; the need and sufficiency for spatial filters whose entries are either zero or an nth root of unity; the problems of target(background contrast and target translation invariance with respect to background; intensity phase- encoding of imagery and its advantages; zero-mean intensity phase-encoding of imagery and its advantages in solving the background problems; and a new edge enhancement (binarization procedure based on phase-encoding and local Wigner transforms. A secondary purpose of this document is to be a user guide to the software created during this effort for the design of spatial filters. A goal of this effort nas been a drastic refinement, extension, and extrapolation of spatial filter design codes in the parallel programming language occam 2 for use in inexpensive transputer based parallel processing systems. Spatial filter design on an occam 2/transputer based system is now approximately 250 times faster than on a FORTRAN/Vax system.