Abstract

The application of a regularization technique to filter synthesis in pattern recognition with synthetic discriminant function filters is presented. The proposed technique uses the stabilizing functional approach for two-dimensional ill-posed problems. Filter synthesis is thus formulated as the minimization of some relevant criteria with specified correlation values for some training input images and limited maximum value of a stabilizing functional. The choice of a particular stabilizing functional to be minimized is related to a priori knowledge regarding the pattern-recognition problem. The analogy between the regularization methods and optimal trade-off filters is also presented and is illustrated with numerical experiments.

© 1994 Optical Society of America

Unconstrained correlation filters

Abhijit Mahalanobis, B. V. K. Vijaya Kumar, Sewoong Song, S. R. F. Sims, and J. F. Epperson
Appl. Opt. 33(17) 3751-3759 (1994)

Generalized synthetic discriminant functions

Z. Bahri and B. V. K. Vijaya Kumar
J. Opt. Soc. Am. A 5(4) 562-571 (1988)

Unified synthetic discriminant function computational formulation

David Casasent
Appl. Opt. 23(10) 1620-1627 (1984)

Optical pattern recognition based on convex functions

Uri Mahlab and Joseph Shamir
J. Opt. Soc. Am. A 8(8) 1233-1239 (1991)

One-dimensional logarithmic harmonic synthetic discriminant function filters for shift-, scale-, and projection-invariant pattern recognition

Jianping Yao and Guy Lebreton
Opt. Lett. 23(7) 537-539 (1998)