Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Optical pattern recognition using dynamic associative memory with orthogonal features extracted by wavelet transforms

Open Access Open Access

Abstract

An optical associative memory is implemented by using orthogonal features extracted through an optical wavelet transformation of a modified Peano scan of a 2-D image. The wavelet transform of 2-D data requires 4-D data representation. This is difficult to perform. By using a modified Peano scan it is possible to reduce the 2-D data to a 1-D vector, thus decreasing subsequent processing to 2-D while preserving much of the proximity relationships. The capacity of associative memories for pseudo-orthogonal vectors is typically about 13% of the total number of neurons. However, this capacity can be increased by using an optical wavelet transformation to extract orthogonal features to be used as input to the optical associative memory. The wavelet functions are typically stored within a fixed medium, producing a static optical storage matrix. By using a dynamically erasable optically addressable storage medium, such as an organic photopolymer, a dynamically reconfigurable storage matrix for wavelet functions is demonstrated.

© 1992 Optical Society of America

PDF Article
More Like This
Optical wavelet matched filters

Yunglong Sheng, Danny Roberge, Taiwei Lu, and Harold Szu
FN1 OSA Annual Meeting (FIO) 1992

Space and Fourier domain holographic associative memory for robust pattern recognition

Taiwei Lu, Andrew Kostrzewski, Hung Chou, and Freddie Lin
MII6 OSA Annual Meeting (FIO) 1991

Hyperspectral Wavelet Transforms for Data Compression and Feature Extraction

James F. Scholl and Eustace L. Dereniak
ThII6 Frontiers in Optics (FiO) 2003

Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.