Optical computing can mean several things: computing with optical hardware, or/and computing of optical signals. Optical signals, in other words: pictures, often consist of many million bits. If pictures have to be processed for example at TV rate, the optical computer has an edge over the digital computer.

A general technique is described for implementing sequential logic circuits optically. The system consists of a nonlinear transducer which provides a two-dimensional array of gates and one or more computer generated holograms (CGHs) to interconnect the gates. The limitations on the number of gates which can be implemented in an optical system is affected by the interconnection method. We describe three interconnection methods and their respective limitations. One method, which is a hybrid of space-variant and space-invariant CGH elements, provides high gate densities and high gate-utilization rates.

Conventional computers suffer from several communication bottlenecks which fundamentally limit their performance. These bottlenecks are characterized by an address-dependent sequential transfer of information which arises from the need to time-multiplex information over a limited number of interconnections. An optical digital computer based on a classical finite state machine can be shown to be free of these bottlenecks. Such a processor would be unique since it would be capable of modifying its entire state space each cycle while conventional computers can only alter a few bits. New algorithms are needed to manage and use this capability. A technique based on recognizing a particular symbol in parallel and replacing it in parallel with another symbol is suggested. Examples using this parallel symbolic substitution to perform binary addition and binary incrementation are presented. Applications involving Boolean logic, functional programming languages, production rule driven artificial intelligence, and molecular chemistry are also discussed.

A simple method for optically implementing digital logic gates in parallel has been developed. Parallel logic gates can be achieved by using a lensless shadow-casting system with a light emitting diode array as an incoherent light source. All the sixteen logic functions for two binary variables, which are the fundamental computations of Boolean algebra, can be simply realized in parallel with these gates by changing the switching modes of a LED array. Parallel computation structures of the developed optical digital array processor are demonstrated by implementing pattern logics for two binary images with high space-bandwidth product. Applications of the proposed method to parallel shift operation of the image, differentia-tion, and processing of gray-level image are shown.

Optical processors performing a large number of multiplications and additions in parallel are ideally suited for matrix algebraic operations, such as vector-matrix and matrix-matrix multiplication. A survey is made of different algorithms and architectures for optical matrix algebraic processors. These architectures are then compared with respect to their hardware requirements and computational efficiencies.

A partially coherent Koehler-illumination imaging system equipped with complementary masks in source and pupil planes can be used to perform image enhancement operations such as directional or non-directional edge enhancement and emphasis of spatial bandpass features. With many objects the use of complementary masks results in high-contrast images. Underlying principles are explained and preliminary experimental results presented.

Long duration signals are recorded as a cohe-rent image in a TV-type raster format on 300 lines, each line being a time,window of duration 1 s with time- bandwidth capacity of 342. A holographic fil-ter produces 3 adjacent images with vertical shifts of 1 line, yielding continuous lines 3 times longer (3 s width) with a sliding time-window stepped by 1 s. An optical multichannel 1 D Fourier transform then performs a time-frequency representation, with a frequency resolution increased by a factor 3. Such a system makes the geometry of coherent light valves more convenient for the multichannel processing of signal with large BT products, for instance in passive Sonar.

A novel approach to electro-optical image manipulation was presented previously. The main advantage of this approach being the mapping of the two dimensional spatial Fourier transform into the temporal domain. This transformation is achieved in real time by using the image modulator whose dynamic optical transmittance at each pixel is given by:

Incoherent optical processing may be anal-yzed by use of the cross-spectral density function, which gives the coherence as a function of wavelength. The cross-spectral density function may be propagated through a general optical system to yield the intensity in the final plane. Analysis of a specific optical processing system shows the effects of component placement, lens aberrations, and illumination bandwidth. In particular, the use of a grating in the input plane introduces a coherence function that allows the optical processing function to take place. Through this key, we may tie together a method of optical processing, a method of image subtraction, and the Lau Effect.

A new unconventional principle of the imaging of the interferometric microscopic objects by using the optical heterodyne detection is proposed and experimentally verified. 1. Principle The optical heterodyne detection is used in imaging devices not only for the photodetector but also for the imaging. A new priciple of the ptical heterodyne is proposed by one of the author '. The optical heterodyne detection with the point reference (local) light source, whose optical frequency is shifted by an optical frequency translater, detects only the point source at the same position of the mirror image of the reference source. Thus by the scanning three-dimensionally the reference point source, an image of the object is obtained. In this system, the frequency translater generates both sidebands to the carrier optical frequency, the beat frequencies from both sideband in the output of the photodetector interferes with each other at the output of the photo-detector, and so the optical phase difference is converted into the phase difference of the resulted beat-frequency signal. Thus the interferometric image of the objects is obtained. In this case, the fringe spacing is half of that of the interferometer.

A fiber-optic display panel is described and demonstrated which magnifies the input image and produces a coherent display on a panel which is black by reflected light.

Low frequency electromagnetic methods for non-destructive testing have been interpreted in terms of the eddy current or inductive phenomenon. In this scenario the flaw distorts the flow of eddy currents, which in turn, results in a change of impedance of the sensing coil. This change of complex impedance can be analyzed in terms of phase an amplitude, and inferences about the depth and size of the flaw made by observing the impedance plane plot on an oscilloscope as the coil is scanned across the flaw. A great deal of operator experience is required, although attempts at automating the decision making process have recently been made. In this paper we explore an alternate view-point; that the change in sensing coil impedance is due to the interference of a reflected electromag-netic wave (by the flaw) with the outgoing wave. This allows us to utilize those properties of wave theory which form the basis for holographic imaging. In this paper we show theoretically and experimentally that such is indeed the case.

In the field of medical imaging the demand for panoramic imaging arises when imaging of human body cavities is of concern, i.e., in the case of endoscopy, since at present the endoscope has to be rotated around its optical axis to get a full 360° image of the entire wall of the body cavity to be investigated. Existing panoramic imaging techniques are not suitable for this purpose, and proposed holographic techniques fall short of doctors' expectations. The scope of this presentation is to review various panoramic imaging modalities, and to introduce a new imaging block which is not only suitable for panoramic endoscopy but which can also be used to re-project the recorded panoramic image onto a ring-shaped screen.

The information contained in a monocromatic wavefield is confined to a spectrum situated on the so called Ewald sphere in the 3D frequency space. We discuss how the Ewald sphere is transformed when the hologram is scaled down and reconstructed at a wavelength different from the recording wavelength using the method of "optimal reference field".

This paper describes a wavefront compensator applicable to microwave imaging antennas that are so large as to be inevitably distorted. Measurements are made of the radiation field due to a single point source located in the general direction in which imaging is performed. Feed-back-controlled circuitry focuses the array upon, the source notwithstanding the unknown a priori distortions within the array. The paper concentrates upon tolerances and system limitations. An experimental example of a two-dimensional radar map with nearly dif-fraction limited resolution is given. The 3 cm wavelength array was 40 m in length and the element position errors exceeded 1 wavelength in all dimensions.

A procedure for accessing the 3-D Fourier space of a conducting or nondispersively scattering object by angular (aspect) and spectral diversity is described. Cost-effectiiie data acquisition in the microwave regime is achieved by substituting spectral degrees of freedom for the more costly angular degrees of freedom set by the number of observation points defining the imaging aperture. A novel target de/Lived Aeietence (TDR) technique is utilized to generate a synthetic phase reference signal emanating from a point on the target achiev-ing thereby a 3-D teri4te44 Fourciek hotognam re-cording arrangement that has many practical advantages. Application of the ptojection-slice theorem, derivable from the multidimensional Fourier transform, to the accessed Fourier space volume is shown to allow the retrieval of 3-D image detail. This can be achieved either tomogra-phically in slices _(or cross-sectional outlines) or as shown here in the form of a projection image of the object scattering function. Examples of micro-wave data acquisition, data normalization for an undesirable range dependent phase term via the TDR method, and image reconstruction are presented for a complex conducting test object. The results shown herald a new generation of high resolution 3-D imaging radars and can be applied in NDE (non-destructive evaluation), biomedical imaging, and remote sensing applications.

In some synthetic aperture systems the data gathered from a single transmitted pulse samples the 3-D frequency space of the object. A 3-D image can be formed by taking a 3-D FT of this data. However, the data are often in a form that allows backprojection techniques to be effi-ciently used to generate the image. We describe an optical hybrid coherent backprojection proces-sing technique that uses a 1-D spatial light modulator, a coherent optical processor, a 2-D detector array, and some simple computer post-processing to produce the image.

We describe a new method that locates radio wave sources by forming images. This approach has potential advantages over conventional direction finding because it shows reflection sources, which lower the accuracy of direction finding; in addition, images show distance as well as direction. The method is to use two sparse, separated arrays of receiving antennas to sample wavefronts. Measured phase and amplitude data are computer processed to form an image. Each array is coherent by means of a phase reference obtained from one of its antennas, but the two arrays are mutually non-coherent. This non-coherence permits multiplying image intensities to improve resolution. An example of experimental results is shown for wave-length 10 meters. The method has been used at wavelengths as long as 50 meters and as short as 33 cm; of course, distinct arrays were used for all these wavelengths.

Fourier transformation is a basic operation in many signal processing applications. A new optical processing approach that is capable of performing in real time cosine and sine transformations on the intensity distribution of an incoherent one-dimensional optical input is introduced. An optical configuration utilizing an achromatic grating interferometer to implement the processing approach is presented. This versatile processing system can be configured to perform operations currently being accomplished by other special purpose optical processors. In addition, it can be used to process signal received directly from incoherent scenes without going through any transducer.

A three-dimensional (3-D) or 4-D signal can be represented as a sequence, i.e. a 2-D set of sectional images. Fourier transform of a whole sequence requires high space-bandwidth transformation capacity and therefore is a appropriate task for coherent optics. The resulting 2-D sequence spectrum is also a sequence representation of the original 3-D or 4-D spectrum. Coordinates and scale of this representation are derived. Several experimental results illustrate this representation and demonstrate the possibilities for multi-dimensional Fourier analysis.

A relation between spatial coherence and source encoding intensity distribution is presented. Since the spatial coherence requirement is determined by the signal processing operation, a strict coherence source may not be needed for the processing. The advantage of the source encoding is to relax the constraints of strict spatial coherence requirement so that the signal processing operation can be carried out with an extended incoherent source. The effects of signal sampling on coherence requirement is discussed. The advantage of signal sampling is to improve the degree of temporal coherence in the Fourier plane so that the signal can be processed in complex amplitude with a broadband white-light source. Finally, experimental demonstrations are included.

We introduce the Wigner Distribution Function (WDF) as a self-windowed Complex Spectrogram (CS) and suggest some methods for the optical generation of the WDF of 2D signals. The resulting WDFs, since they are 4D functions, are represented as sectional images displayed either in parallel or as temporal sequences. We show some typical results obtained from our coherent-optical WDF-processor using the latter display technique.

An extension of [1] to measurement of vibrating modes of an object using holographic spectroscopy is given. The calculation of the average irradiance for a random vibrating object considering different conditions of the reference beam is described. The details of the experimental setup based on a phase locking principle is proposed.

Optical processing systems which perform linear transformations on image data at high rates are ideal for image pattern recognition systems. As a result of this processing capability, the linear opera-tion of matched spatial filtering has been explored extensively for pattern recognition. For many practical pattern recognition problems, however, multiclass filtering must be used to overcome the variations of input objects due to image scale changes, image rotations, object aspect differences and sensor differences. Hester and Casasent have shown that a linear mapping can be constructed which images all the class elements of a multiclass set into one out-put element or value. This special multi-class filter concept is extended in this paper to show that a subspace of the multi-class set exists that is invariant with respect to the multiclass mapping under linear operations. The concept of this in-variant space and its generation is detailed and a single example given. A typical optical processing architecture using these invariant elements as filters in an associative pattern recognition system is also presented.

The Wigner Distribution Function (WDF) is an excellent tool for characterizing time varying signals. Several approaches have been recently suggested for optical implementation of Wigner Distri-bution Function of signals. In this paper, we report on our simulation efforts to use WDF for signal and image classification. We also relate WDF to other conventional time-frequency representa-tions of a signal such as Ambiguity Function. We develop analytical results quantifying the performance of WDF in detection problems. These analytical results are compared to the experimentally observed results. Signal parameter estimation methods that use WDF are also introduced. Extensions of WDF approach to higher dimensions (as in the case of images) are briefly outlined.

In this paper, we present a theoretical model for the transmission function of a biological cell. Three different cases have been considered: a) Real transmission function, b) Pure phase Object, c) Mixed case where both, real and complex terms are present in the transmission function. We have studied in all those cases the response in the Fourier plane. The comparison of the models proposed above allow us to conclude that the most correct treatment would come from c). Even if the cytoplasm gives the most important contribution, neither the nucleus nor the membrane can be neglected.

The finite aperture of any physical imaging system eliminates the high spatial-frequency components of the object from appearing in the image. The lack of high frequency detail results in a loss of resolution in the observed image. It has been shown that, for an object of finite extent, an exact restoration of the object from a DL image is possible. However, numerical implementation of the DL image restoration process is highly unstable in the presence of measurement noise. In the dual of the image restoration problem, the extrapolation of a finite segment of the DL (i.e. spatially limited) image data in the presence of measurement noise is performed. It has been found that the imposition of a priori constraints, such as a non-negativity of the estimate, will stabilize the restoration process. In this paper, we employ optimal data fitting techniques that uses linear programming (LP) for optimization. Results of numerical experiments are presented to illustrate the efficacy of this approach.

Rotation blur cannot be described by a convolu-tion in x-y space, as, e.g., linear motion blur. Hence inverse filtering (deconvolution) in x-y is not possible, the system to be inverted is shift-variant. By modelling the recording process (e.g. on photographic material) in three dimensions (x-y and t), however, the blurring process can be described as a convolution and a multiplication. The 3D functions involved can be handled by a coherent-optical processor by means of sequence representations. Hence deblurring becomes essentially a sequence deconvolution. The sequence deconvolution point response is discussed and experimental re-sults for blur simulation and deblurring are shown.

By projecting 16mm films of computer-generated images onto a thin wall of live steam, an illusion is created of artificial 3-D. The effect of visual scale and distortion of depth clues make the visual property of this phenomenon unusual. The water droplets also produce a "rainbow" effect of adding color to black and white film.

The coherent optical filtering techniques provide a general concept for the classification of patterns. This paper describes the design and testing of a hybrid optical-digital image processing system and the development of methods for a statistical expansion of the correlation signals. A conventional correlation signal intensity measurement is in most of the applications not sufficient. Six different algorithms for correlation signal evaluation are investigated. A feature reduction is achieved by multivariate analysis. For alpha-numeric patterns distored by binary random noise, rotation, scaling and shearing high classification results have been optained.

This paper describes the development of a hybrid Opto-electronic System that allows for the automatic extraction of displacement information from a sequence of holographic interferograms. The technique utilizes opto-digital image processing to determine the passage of fringes fOr a set of surface points simultaneously. Several of the notions included permit the fast and accurate quanti fication of displacement for large volunns of test data. The implementation of this technique will allow for the application of this relatively sophisticated non destructive testing method to the solution of traditional engineering mechanics problems.

We report the fabrication of n-channel Metal-Oxide-Semiconductor Field-Effect-Transistors (n-MOSFETs) in laser recrystallized silicon films on 'a lithium tantalate (LiTa03) substrate with an intervening silicon dioxide layer. Constraints on film thickness and laser scan conditions are described which lead to melting of the silicon film without damaging the substrate. Electrical characteristics of the n-MOSFETs are presented, and the devices are shown to exhibit an electron channel mobility of 50 cm²/V-sec. Theoretical operation of the n-MOSFETs as integrated photodetectors of substrate light is derived and compared to observed behavior.

Digital optical processing (DOP) was conceived to encompass the advantages of both electronic and optical processors, which are parallelism, flexibility, and high accuracy. In this paper we discuss the concept of parallelism, how it applies to DOP differently than to Electronic Parallel Processing, and other potential advantages in using DOP. A PLZT device will also be described which can perform a series of logic or memory operations. From several of these PLZT devices a DOP is constructed to illus-trate some of its programmability features.

Optical approaches to solving the Ax = b problem have suffered from four difficulties: (1) an inability to handle the problem for nonsquare A, (2) the necessity of insuring convergence for nonsingular A, (3) the inability to handle a singular A, and (4) inaccuracies due to an ill conditioned A. We show that these problems can all be solved or mitigated by singular value decomposition (SVD). An accurate approach to optical SVD is shown.

Light propagation through the atmosphere is disturbed by atmospheric turbulence. This limits the high angular resolution, in astronomical imaging. Active optics is a method to overcome this problem. It allows a real-time optimization of the resolving power. An active mirror was developed which consists of an electrostatically deformable membrane. For an atmospheric tilt compensation of the wave-front the mirror housing is in a gimbal mount with piezo-electric actuators. With this active mirror device and a multi-microprocessor control unit the stabilisation of the star-speckle pattern positions and the deconvolution of the speckle patterns are possible. The compensating phase distribution is generated by an expansion of the turbulence phase distortions into modes of a set of basis functions, i.e. Zernike-polnomials or Karhunen-Loeve-func-tions. By applying a modal control concept to the adaptive optical system, the electrodes of a mem-brane mirror are controlled in parallel with com-pensated cross-talk. The coefficients of the ap-proximating functions, each of which corresponds to a mode on the mirror surface, are fed back to the actuators by a modal control matrix.

The relationship between phase and amplitude in matched filters is examined. From image processing work, we know that the phase information is significantly more important than amplitude inform-ation in preserving the features of a visual scene. Is the same thing true in the case of a matched filter? From previous workl we know that a pure phase correlation filter can have an optical effi-ciency (Horner Efficiency) of 100% in an optical correlation system. The trade-off for this is in the discrimination ability of the filter. We examine this relationship in the case of alpha-numeric characters using a computer simulation. Phase-only and amplitude-only filters are analyzed. Three-dimensional plots of the autocorrelation and crosscorrelation functions are presented and discussed.

Our unified synthetic discriminant function (SDF) filter synthesis technique using the correlation matrix of the image training set is reviewed. Four different synthetic discriminant functions for intra-class recognition, inter-class discrimination and both intra and inter-class pattern recognition are considered. All techniques proposed are appropriate for object identification, location and classification in the presence of 3-D geometrical distortions in the input object. Initial results obtained on a set of four different classes of infrared ship imagery are presented. Excellent performance (over 90% correct classification) was achieved.

Laser Diodes are becoming increasingly attractive for application to time integrating and systolic optical computing architectures where large bandwidths and coherent processing are required. This paper reports the results of a study to determine the interferometric performance of laser diodes when used in a pulsed mode and where the interference effects of many sequential pulses are integrated temporally. The results indicate qat fringe visibility is high for as many as 104 integraged pulses if the OPD in the processor is less than 5 mm and the laser pulse width is less than 50 nsec.

Optical image correlators are presented that use acousto-optic and charge-coupled devices as the input and output transducers respectively. Experimental results are presented and the applicability to pattern recognition of a non-linear pseudo-correlation that can also be conveniently computed with these processors, is discussed.

The design, error analysis, component accuracy required, computational capacity, data flow and pipelining, plus the algorithm and application all seriously impact the use of optical systolic array processors. This paper provides initial remarks, results, examples and solutions for each of these issues.

A method of modifying the shape of the modu-lation transfer function of non coherent optical systems based on simultaneous subtraction in real time of the responses of two systems having different pupils is described. When one-dimensional images are treated, the problem is relatively simple, since the second dimension can be used solely for encoding and thus few limitations are posed on the apertures of the optical system. In this paper the design of two-dimensional apertures providing the necessary emphasis of the MTF components in all spatial directions is presented. A circular aperture for one optical system and an annulus one for the second is optimal. Criteria for the determination of the characteristics of these apertures are derived.