Classification of real-time X-ray images of randomly oriented touching pistachio nuts is discussed. The ultimate objective is the development of a system for automated non- invasive detection of defective product items on a conveyor belt. We discuss the extraction of new features that allow better discrimination between damaged and clean items (pistachio nuts). This feature extraction and classification stage is the new aspect of this paper; our new maximum representation and discriminating feature (MRDF) extraction method computes nonlinear features that are used as inputs to a new modified k nearest neighbor classifier. In this work, the MRDF is applied to standard features (rather than iconic data). The MRDF is robust to various probability distributions of the input class and is shown to provide good classification and new ROC (receiver operating characteristic) data. Other applications of these new feature spaces in robotics and face recognition are also noted.

The high data transfer rate achievable in page-oriented optical memories demands for parallel interfaces to logic circuits able to process efficiently the data. The Optically Programmable Gate Array, an enhanced version of a conventional FPGA, utilizes a holographic memory accessed by an array of VCSELs to program its logic. Combining spatial and shift multiplexing to store the configuration pages in the memory, the OPGA module is very compact and has extremely short configuration time allowing for dynamic reconfiguration. The reconfiguration capability of the OPGA can be applied to solve more efficiently problems in pattern recognition and digit classification.

In this paper we show how to use holographic photon echoes for the implementation of a variety of optical processing functions, including scanners, spectrum analyzers, time- integrating correlators, folded spectrum analyzers, ambiguity function processors, image sequence correlators, and folded image raster correlators. The combination of optical coherent transients operating as spatial-spectral holograms with acoustooptic deflectors and electrooptic modulators allows a variety of optical architectures to be implemented with substantially enhanced performance and functionality beyond that achievable by these technologies individually. We show how to utilize the basic architectures presented here as building blocks of more powerful and complex real-time optical processing systems.

Cross-connect switching is a common switching architecture for telecom and datacom applications. Large bandwidth O-E interface devices have recently been made commercially available. Small scale fast electronic switches and large scale optical interconnect circuits can be effectively used for handling large bandwidth O-E cross-connect switching. In this paper, we show two packaged and connectorized optical interconnect circuits. The first one is a 100 X 100 channel guided-wave circuit fully compatible, through MT array connectors, to O-E interface devices, such as Motorola OPTOBUS^TM or Simens PAROLI^TM chips. The second one is a more scalable architecture which is a hybrid of free- space and fiber circuits. For demonstration purpose, a 256 X 256 channel hybrid circuit is shown. Key parameters, such as insertion loss, cross-talk, and bit-error-rate of these interconnect circuits are presented. Transmission and routing of video data are performed to demonstrate interconnect quality of various data links. Scalability of these demonstrated circuits to larger sizes are speculated.

An optical interconnection technique using photorefractive segmented waveguides is reviewed. The photorefractive segmented waveguide consists of many localized high refractive-index regions that are fabricated by illuminating a focused laser beam. The waveguide interconnections can be adaptive by changing the spatial arrangement of the high refractive-index regions. Fabrication results of straight, curved, and Y-branch waveguides are presented. In the straight waveguides, the transmitted power of a guided beam as a function of the period of segmentation and the dark decay time are measured. The tolerance due to fabrication error is also investigated experimentally. Waveguide structures can be optically modified to implement adaptive interconnections. As an example, a curved structure was transformed into a Y-branch structure by optical illumination. We numerically show that a novel structure of a segmented waveguide can work as a waveguide, where average refractive index of the core is the same as that of surrounding material.

We present an adaptation of the BEAMTAP (Broadband and Efficient Adaptive Method for True-time-delay Array Processing) algorithm, previously developed for wideband phased array radars, to lower bandwidth applications such as sonar. This system utilizes the emerging time or wavelength multiplexed optical hydro-phone sensors and processes the cohered array of signals in the optical domain without conversion to the electronic domain or digitization. Modulated signals from an optical hydro-phone array are pre- processed then imaged through a photorefractive crystal where they interfere with a reference signal and its delayed replicas. The diffraction of the sonar signals off these adaptive weight gratings and detection on a linear time- delay-and-integrate charge coupled device (TDI CCD) completes the true-time-delay (TTD) beamforming process. Optical signals focused on different regions of the TDI CCD accumulate the appropriate delays necessary to synchronize and coherently sum the acoustic signals arriving at various angles on the hydro-phone array. In this paper, we present an experimental demonstration of TTD processing of low frequency signals (in the KHz sonar regime) using a TDI CCD tapped delay line. Simulations demonstrating the performance of the overall system are also presented.

We present a time and space integrating optical architecture for multi-layer finite impulse response neural networks (FIRNN). The proposed architecture is capable of forward propagation and on-line learning in the form of backward propagation. FIRNNs are first presented and analyzed. From the analysis it is observed that the implementation of FIRNNs requires the calculation of temporal convolutions, which inspire the use of time-integrating and space- integrating optical architectures. A novel device is proposed for the space-integrating architecture, based on the use of a rotating volume hologram. Initially, two single-layer architectures based on the space integrating and time integrating architectures are presented, leading to the multi-layer architecture, which uses a combination of both architectures, folding them together in such a way that all the operations of the order O(N³) are performed optically and only the less computationally intensive operations are performed electronically.

The performance of linear and nonlinear processors (filters) for image recognition which are I_p-norm optimum in terms of tolerance to input noise and discrimination capabilities. These processors were developed by minimizing the I_p- norm of the filter output due to the input scene and the output due to the noise. The performance of the I_p-norm optimum filters is tested by measuring the average peak-to- sidelobe ratio of the output of the filters for different values of p. In addition, the performance of these filters is tested further by placing a target in a scene containing additive noise and realistic background scene. For the images presented here, the filters detected the target in the presence of additive noise and realistic background scene.

Litton PRC and Litton Data Systems Division are developing a system, the Imaged Document Optical Correlation and Conversion System (IDOCCS), to provide a total solution to the problem of managing and retrieving textual and graphic information from imaged document archives. At the heart of IDOCCS, optical correlation technology provides the search and retrieval of information from imaged documents. IDOCCS can be used to rapidly search for key words or phrases within the imaged document archives. In addition, IDOCCS can automatically compare an input document with the archived database to determine if it is a duplicate, thereby reducing the overall resources required to maintain and access the document database. Embedded graphics on imaged pages can also be exploited; e.g., imaged documents containing an agency's seal or logo can be singled out. In this paper, we present a description of IDOCCS as well as preliminary performance results and theoretical projections.

We present an efficient segmentation algorithm to discriminate between different materials, such as painted metal, vegetation, and soils, using hyperspectral imagery. Most previously attempted segmentation techniques have used a relatively small number of infrared frequency bands that use thermal emission instead of solar radiation. This motivated the use of hyperspectral (or multispectral) imagery for segmentation purposes taken at the visible and near infrared bands with high spectral dimensionality. We propose a segmentation algorithm that uses either a pattern- matching technique using the selected band regions or a principal component analysis method. Segmentation results are provided using several hyperspectral images. We also present a band-selection process based on either pairwise performance evaluation or a band-thickening method to select the particular band regions that contain important band- value information for segmentation. A hyperspectral data set that contains a number of spectral band-value curves collected from eleven hyperspectral images is used as an evaluation data set for the band-selection process.

Frequently in character recognition, there is a need to correlate or convolve purely real 2D inputs and real 2D filters. We propose two optical correlators. The first, basic real-input-real-filter optical correlator, adds x and y mirror images adjacent to the image to generate real and even images. This avoids the need for an offset reference to capture phase. Therefore, we have greater diffraction efficiency and a reduction in filter LCD pixel size. The second, Hilbert transform basic real-input-real-filter optical correlator, uses a Hilbert transform by masking in the filter plane to reduce the number elements in the filter LCD to that of the number of pixels in a filter image. An optical system can generate the mirror images in real time so that both the input LCD and the filter LCD can have only the same number of elements as their respective number of image pixels. We show how a spatial filter can be used to remove the intensity offset produced in the Hilbert transform and restore shift invariance if required. Finally, if desired, some rotation invariance can be achieved by overlapping the mirror images with the basic correlator. Advantages and disadvantages of the proposed correlators are discussed.

We investigate the performance of morphological correlation under different illumination conditions. The morphological correlation is shown to be invariant to uniform input-image illumination when the input-image illumination is higher than that of the reference. Accordingly, illumination- independent pattern recognition can be realized using morphological correlation provided the reference is multiplied by a proper number less than unity or the input image is multiplied by a number greater than unity before threshold-decomposition. In addition, the correlation peak tends to get broader when illumination of the input is different from that of the reference. Computer simulation results are provided.

We have demonstrated a color recognition system composed by a novel, self-organized optical neural network system that includes genetic algorithms along with back propagation schemes, which make it possible that the system avoids the local minimum problems and make the learning processes faster and better. Our system is composed of a color liquid crystal display panel (LCD), bistable semiconductor lasers, photo-diodes, and liquid crystal light projector (LCP). The LCD weighs the intensities of light that passes through it and works as synapses in the neural network. The optical bistable semiconductor lasers originate the optical sigmoid functions and serve as threshold processing units. Using these devices can simplify the configuration of the optical neural network system. The color of light emitted from the LCP will be recognized by the neural network system. A monochromatic light beam generated by the LCP is illuminated on all over the LCD surface displaying the colored boxes in the three primary components. Thus, the light beam is weighed when it passes through the boxes on the LCD. As a consequence, we have achieved a novel, simplified color recognition system using the genetic algorithms for self- organization of the optical neural network. The unique feature of this system is to make use of the genetic algorithms and the back propagation at the same time to derive selectively the merits from these two methods. By this system, more naturalized color recognition like human will be performed, being able to distinguish the colors under different conditions of environment, e.g. lightening conditions, surface conditions of colored material, etc.

We propose a new technique of correlation filter design for optical correlator. For detection and classification of complex patterns, ability of single optical correlator is not enough. To achieve high performance of classification, a multiple correlator is suitable in respect of flexibility in correlation filter design. We attempt to design a set of correlation filters for use in multiple optical correlators. As the target, we select road signs. In real scene, the scale and aspect ratio of road sign are dependent on the distance and angle of observation. In addition, many kinds of signs are used. Therefore the correlation filter set has to be designed as to have distortion invariance to adapt to the change of aspect and the variation of road signs. We apply the technique of multiple-object correlation filter, such as the synthetic discriminant function, to the design of correlation filter set, in order to obtain the necessary invariance. We calculate 180 filters in order to detect and classify 15 kinds of Japanese road signs in real scene. Computer simulation result shows that the combination of multiple optical correlator with the correlation filter set can indicate high performance of pattern detection and classification.

The spatio-temporal distribution and evolution of wave fields is measured by the acquisition of optical video image sequences. An experiment at the wind wave flume of the University of Hamburg was performed to measure small-scale processes at the water surface. From the image sequences, image spectra are calculated by a 3D Fast Fourier Transformation. This spectral description is limited to stationary and homogeneous wave fields and therefore other techniques which are designed for local scales are applied.

We consider the recently proposed double phase encoding system, and we analyze the influence of a perturbation of the coded image on the decoded image. Since in optical implementation, the intensity of the decoded image is measured at the output of the decoding system, we express the signal-to-noise ratio (SNR) on the intensity decoded image as a function of the amplitude SNR of the complex coded image. We show that there exists a simple relation between these two SNR's, and that this relation is quasi- independent of the nature of the coded image perturbation. The results presented in this paper could help characterizing the precision level of the components of the optical decoding system necessary to reach a pre-defined quality level of the decoded image.

An encrypted holographic memory based on double random phase encoding and angular multiplexing is presented. A 2D digital image is encrypted by two random phase codes located at the input and Fourier planes. The input digital image is displayed on a liquid crystal display. In a photorefractive LiNbO₃:Fe crystal, Fourier-transformed patterns of encrypted images are stored holographically by using angular multiplexing. Experimental results of encryption and decryption of digital data are presented.

A novel architecture for the double random phase encoding optical encryption system using a joint transform correlator is proposed. In the proposed method, the joint power spectrum of the image to be encrypted and the key code is recorded as the encrypted power spectrum. Owing to the joint transform architecture, the alignment of the image and the key code is robust. Computer simulations using a computer generated hologram is presented to confirm our proposed system.

An optical security system using a random selected binary phase code is proposed. A random binary phase encoded light is incident on a holographic recording material and writes gratings in the material with a reference plane wave. The recording material can be regarded as an optical lock while the phase code acts as the key. As the pixel number is 5 X 20, the decryption probability is less than 10^-10 when arbitrary phase code is used for decryption. The decryption probability decreases dramatically as the pixel number increases. Furthermore, the security system performs good tolerance to data loss of the phase code.

An idea for security verification of identification such as passports, credit cards, and others in terms of difficulty of reproduction is proposed. The polarization information is used for security verification. As the state of the polarization encoded cannot be seen by an ordinary intensity sensitive device such as a CCD camera, it cannot be copied. For optical validation system, a joint transform correlator is used. Computer simulations and optical experimental results are shown to confirm our proposed method.

In this paper we present some results on the research line of real-time correlation applied to optical pattern recognition. We present a real time optical correlator that uses two twisted nematic liquid crystal spatial light modulators (SLM). This kind of SLM produces phase and amplitude coupled modulation. Phase modulation is useful for the implementation of phase only filters (POF), but the coupled amplitude modulation affects its performance. We present a model to evaluate the response of the phase-only filter (POF) implemented on a modulator with a restricted modulation and we show results on the optimal implementation of POF of this restricted coding domain. The phase modulation is also used for the implementation of the scene by means of a phase-encoding algorithm. The performance of the phase-only filter has been improved for several performances criteria by the design of binary amplitude masks. We present a technique for the implementation of these optimized binary amplitude phase-only filters using phase-mostly SLMs. Results obtained for the optimization of different criteria are presented. Finally, we present results on the application of these methods to color pattern recognition problems by a multichannel correlation.

Double random phase encoding technique is a valuable and effective method for optical image encryption. However, a precise alignment is required when optical setup is performed. In this paper, we investigate the shift tolerance property of the technique. The theory of the robustness to data loss of the encrypted image is proposed. According to the theory, we propose a simple and novel method to improve the shift tolerance of the decrypting phase mask. Both theory and computer simulation are presented.

In this paper effective differential algorithms of the correlation analysis (CA) of short time series in real time oriented on multi-process and neural systems have been analyzed. According to this analog signal representation delta modulation (DM) format as well as the correlation analysis of recurrent algorithms have also been considered. Particular attention has been paid to the use of sign delta modulation (SignDM) in CA proposed by the authors. Correlation analysis in SignDM format was illustrated with the results of computer simulations. The algorithms that were worked out are regular and particularly useful in neural systems.

The hybrid optical-digital joint transform correlator (HJTC) experimental setup was created. The minimum focal length of a Fourier lens in the joint Fourier transform processor (JFTP) was found. The procedure of the transformed phase mask (PM) identification in the HJTC was represented. A several modifications of the setup with different effective focal length of the JFTP were used for investigations of transformed PMs. The peak-to-noise ratio (PNR) for each correlation peak in the first diffraction order of the output correlation signal was selected as the objective criterion of the transformed PM identification performance. The performance of the transformed PMs identification in the HJTC was studied. The dependencies PNR versus effective focal length of a JFTP were analyzed. The influence of the CCD-camera recording aperture relative dimensions on degradation of correlation peaks was studied.

In this paper, we combined the multivariate discriminant analysis into an optical correlator and thus improved the distortion-invariant recognition ability of the processor. In this approach, a set of eigenimages are first extracted from a large number of training images including various distortions by using the K-L transform and then are used as the reference images in the optical correlator. The correlation results between the testing image and the set of eigenimages construct a feature space, on which the multivariate discriminant analysis is performed. As a result, a set of low dimensional discriminant vectors representing each image in the training set will be obtained and saved in memory during the training process. When any testing image with unknown membership inputs, it will be processed with the same operations and gets its discriminant vector. Using the simple minimal distance rule, the testing image can be classified into a group whose discriminant vector approximates that of the testing image most. Because the images in the training set are selected to representing all the typical distortions in each group, the algorithm can deal with the distortions to a large extent.

Based on the theory of volume holographic associative storage in a photorefractive crystal and that of binary optics, a compact dual multichannel optical wavelet transform processor is proposed and constructed. Both wavelet correlation and wavelet transform can be complemented by the system. Multi-pattern channels are achieved by the inherent parallelism of volume holographic storage. Angle multiplexed holograms of wavelet filtered pattern images are recorded in the crystal Multi-wavelet channels are accomplished by a Dammann grating, which is a binary optical element for spectrum duplication. The grating is adopted to generate a set of channels with different wavelet filters. Wavelet correlation peaks in different wavelet channels are synthesized to improve the recognition accuracy by multiplication pixel by pixel. Wavelet transform results in different wavelet channels are stored in the crystal and can be restored for recognition or segmentation. The application of the system in human face recognition is studied.

Interferometric 3D imaging based on measuring the space- variant cross-spectral density propagated in free space is reviewed. The spatial coherence function such as the cross- spectral density or the mutual intensity conserves the total 3D information as propagating in free space. The present method is on the basis of the propagation law of the cross- spectral density described in the Fourier domain. By using this law, the cross-spectral density across an arbitrary transverse plane is retrieved from the cross-spectral density measured on the observation plane. A series of retrieved cross-spectral densities composes the 3D intensity distribution. An experimental retrieval of the intensity distribution is demonstrated by using uncorrelated two point sources. A similarity of a point spread function of the present interferometric imaging system to an ordinary image forming system is also discussed.