Publication
Optics Communications
Paper

Neural networks for binarizing computer-generated holograms

View publication

Abstract

A Hopfield type neural network was applied to compute synthetic, binary holograms. This iterative neuron algorithm minimizes reconstruction errors in amplitude and phase. About twenty iteration steps are sufficient for convergence, each of them has the computational complexity of a FFT. In addition, the neural net is flexible to include a variety of optimization strategies. © 1991.

Date

Publication

Optics Communications

Authors

Share