| Research Support Tool |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Context
A non-zero order joint transform correlator for space-variant pattern recognition
Title:
A non-zero order joint transform correlator for space-variant pattern recognition
Archive:
Eprints@IIT Delhi
Author(s):
Unnikrishnan, G
Joseph, Joby
Singh, K
Joseph, Joby
Singh, K
Date:
2005-09-01
Abstract:
We propose a space-variant correlator using a joint transform correlator JTC architecture. The reference and target
images of a JTC are modulated by a random phase function which is the sample of a stationary white process ergodic in
mean and variance. The degree of shift-variance can be varied by changing the variance of the random phase function. The
joint power spectrum JPS is a sinusoidally varying function with a spatially nonvarying average value. This property is
advantageous in optical implementations when the JPS is captured by devices such as CCD with a limited dynamic range. The zero-order and the correlations approach delta functions in the correlation plane. By subtracting a constant value equal to its spatial average from the JPS, the undesirable zero-order can be eliminated.
images of a JTC are modulated by a random phase function which is the sample of a stationary white process ergodic in
mean and variance. The degree of shift-variance can be varied by changing the variance of the random phase function. The
joint power spectrum JPS is a sinusoidally varying function with a spatially nonvarying average value. This property is
advantageous in optical implementations when the JPS is captured by devices such as CCD with a limited dynamic range. The zero-order and the correlations approach delta functions in the correlation plane. By subtracting a constant value equal to its spatial average from the JPS, the undesirable zero-order can be eliminated.
Index terms:
Discipline(s):
Optical pattern recognition
Subject(s):
Joint-transform correlator; Space-variant pattern recognition
Method/Approach:
Coverage:
Publisher:
Contributors:
Source:
Language:
en
Relation:
Type:
Article
Format:
1744499 bytes
application/pdf
Copyright Information:
Browse Archive