N. Konforti

Phase-only modulation with twisted nematic liquid-crystal spatial light modulators

It is shown that twisted nematic liquid-crystal spatial light modulators behave as phase-only modulators when operated below the conventional optical threshold. Thus such devices, when operated in a reflection mode, behave as spatial amplitude modulators when used between crossed polarizers above the optical threshold; they behave as phase modulators when used between parallel polarizers and operated below that threshold.

Shift and scale invariant pattern recognition using Mellin radial harmonics

Abstract Design of matched filters containing a single radial harmonic of an object are shown to provide shift and scale invariant correlations. The theory of the expansion, using Mellin radial harmonics, is analyzed and tested experimentally for some binary optical targets. Cross-correlation pattern recognition in the presence of scale and shift invariances has been successfully demonstrated.

Wavelength tunability in a Nd-doped fiber laser with an intracavity polarizer

When an intracavity polarizer is introduced into a Fabry-Perot Nd-doped fiber laser, wide wavelength tunability is achieved either by rotating the polarizer or by varying, the fiber birefringence with an in-line polarization controller. The intracavity polariser also produces considerable line-narrowing. A model based on birefringence dispersion is proposed.<<ETX>>

Dynamic optical interconnections

An optical configuration is described that uses an addressable reflective component (a liquid-crystal light valve) permitting free-space interconnections between points located on the same substrate. The scheme is attractive for generating programmable interconnections between electronic elements located on the same chip or wafer, in particular since it is highly resistant to system misalignment. Experimental results indicating a fan-out capability, which is desirable for clock signal distribution, are presented.

Improved rotation or scale invariant matched filter

Optical correlation schemes based on circular or radial harmonics have been recently suggested for performing pattern recognition in the presence of rotated or scaled versions of the object. Those schemes, however, are based on using a single harmonic at one time. In this paper we present two approaches that overcome this restriction.

Complex reference-invariant joint-transform correlator

The joint-transform correlator operation has been extended to include complex reference images. Such amplitude and phase images are common when single harmonic images (circular, radial, logarithmic, etc.) are used in correlation setups. The analysis of a rotation-invariant pattern-recognition system using joint-transform correlator principles is described, followed by experimental results that can be obtained in real time.

Real-time optical generation of circular or Mellin radial-harmonic filters

Invariant pattern-recognition systems use spatial filters that are, in most cases, computer-generated holograms. A novel approach for the optical generation of circular-harmonic or radial-harmonic filters is presented, together with experimental results.

Four-quadrant optical matrix-vector multiplication machine as a neural-network processor.

Optical processors for neural networks are primarily fast matrix-vector multiplication machines that can potentially compete with serial computers owing to their parallelism and their ability to facilitate densely connected networks. However, in most proposed systems the multiplication supports only two quadrants and is thus unable to provide bipolar neuron outputs for increasing network capabilities and learning rate. We propose and demonstrate an opto-electronic four-quadrant matrix-vector multiplier that can be used for feed-forward neural-network recall and learning. Experimental results obtained with common commercial components demonstrate a novel, useful, and reliable approach for fourquadrant matrix-vector multiplication in general and for feed-forward neural-network training and recall in particular.

Scale Invariant Pattern Recognition

Design of matched filters containing a single radial harmonic of an object are shown to provide shift and scale invariant correlations. The theory of the expansion, using Mellin radial harmonics, is analyzed and tested experimentally for some binary optical targets. Cross-correlation pattern recognition in the presence of scale and shift invariances has been successfully demonstrated. Two approaches for improving t he correlation peak are presented.

Invariant Pattern Recognition: Towards Neural Network Classifiers

From the many studies on invariant pattern recognition, the coauthors have selected single harmonic filters [1,2]. A real-time optical implementation of such filters was demonstrated on photographic input images [3], but its 10-F length yielded a poor signal-to-noise ratio, making it unusable with the limited dynamic range and flatness of a light valve for real-time video input. For real applications, the other difficulty is the usual requirement of all invariances simultaneously: scale and projection can be processed in a single filter [1], but a pre-processing is necessary for the rotation before any object recognition, since the projection and scale invariant filter do not work on rotated objects.