The applications of Dyed Gelatin and Polyvinyl alcohol for non-Fourier image processing are described. Examples of contrast reversal, edge enhancement, character recognition and addition and subtraction of images using Dyed gelatin films are presented. Dyed Polyvinyl alcohol has been used in a Optical Correlator for the detection of Cross Correlation peak which can serve as an input for a neural network.

Winner-takes-all dynamical behavior of transverse structure in a large aperture dye laser with intracavity saturable absorber on bacteriorhodopsin controlled by bleaching and coloring external beams is demonstrated both theoretically and experimentally. As a result of mode competitive interactions in the multistable dye laser the strongest component of multimode structure incoherently initiated by uniform external bleaching beam suppresses the rest modes and the final stage of transverse structure evolution is the single-mode regime of operation.

Dynamic spatial and temporal properties of accumulated stimulated and modified stimulated photon echoes are considered. Possible applications to optical information processing systems for the temporal compression of the recorded information, the erasure of information, as well as for the correlation analysis of optical images are discussed.

A method is proposed for recording and development of optical information on photothermoplastic (PT) media. The method is based on using radiation of an YAG:Nd laser. Specialized lasers and recording devices on their basis have been developed which ensure obtaining nanosecond (about 30 ns) radiation pulses in the visible range ((lambda) equals 0.532 micrometers ) to be used for exposure in the presence of infra-red background radiation ((lambda) equals 1.064 micrometers ) of quasi-stationary lasing of large duration which is used for thermal processing of the medium. The recording unit proposed makes it possible to realize a multi- stage PT recording process including exposure and thermal processing of the medium by one laser shot.

We proposed to periodically refresh the images stored in dynamic holographic memories by using an opto-electronic feedback loop. Readout of the images without loss of information is experimentally demonstrated in a photorefractive LiNbO₃ sample.

A new all-optical regenerator of the soliton pulses for a fiber memory is proposed. Semiconductor structures based precise recovery of basic parameters of the soliton pulses such as amplitude, duration, shape of envelope, frequency chirp and carrier wavelength are considered and some numerical estimations are presented.

Nonstationary energy exchange between reading pulse and reconstructed waves is analyzed as a function of their energy and spectral characteristics, as well as of their small-scale instability. It is shown that for small values of spatial frequencies, (kappa) ₁, in transverse structure of object wave, phase-locking of reconstructed waves is observed which stabilizes the energy exchange. At the same time the efficiency of wavefront reconstruction is substantially influenced by the instability of the reading pulse with respect to small-scale perturbations of its transverse structure. With increasing (kappa) ₁ the above process depends also on the effect of phase accumulation. In optically dense media the suppression of instability is not observed, but there are special conditions defined by the value and sign of detuning of reference pulse carrier frequency from resonance, under which this instability grows slowly.

The photoinduced transformation of molecular pairs in solid polymer compositions with polycyclic hydrocarbons have been found to result in the enhancement of luminescent and volume phase optical pictures. The effects are caused by electron energy transfer process, reversible photodimerization and structural relaxation. Proposed mechanisms provide possibility of direct reversible holographic and luminescent optical recording with sensitivity higher than 10 cm²J^-1 and even optical information processing based on optical bistability.

Developed method of local thermoplastic recording with IR-laser heating provides qualitative phase holograms registration. Fourier-hologram registration in frequency plane with low spatial frequency filtration is obtained. The possibility of linear regime of the holograms recording on photothermoplastic material is shown. Also, the holograms recording on photothermoplastic material with multiplexing by using developed simultaneous method is demonstrated. Using this approach holographic photothermoplastic disk with two-dimensional fourier-holograms registration is realized. It supplies high storage capacity and information transfer rate. The main advantage of holographic disk with thermoplastic recording using in such implementations as correlator, optical neural network and associative memory are shown.

As a parallelism of digital computers increases, the limitations associated with interconnecting a large number of processors becomes a greater and greater constraint on system performance. For example, as we pack more and more processors of a given size together in a single machine, naturally the physical dimensions of the machine must grow, and with that growth comes an increase in the maximum time delay experienced in communicating between the most distant processors in the array. To a degree that depends on the algorithm being executed and its communication requirements, that communication latency between different parts of the machine ultimately poses a limit to the speed with which the machine can solve problems. Our purpose in this paper is to discuss some of the fundamental aspects of the problem described above. We consider in what follows two different cases: (1) all interconnections are optical, and (2) all interconnections are electrical. Of course with a hybrid set of interconnects (i.e. part optical and part electrical), better performance can be achieved. However, we do not consider hybrid strategies in this paper.

Synthetic diffractive optical elements offer unique wavefront transformation capabilities, and, consequently, the means to solve many optical interconnection problems encountered in parallel all-optical and electro-optical processors. We describe the design and fabrication of a wide range of fixed space-invariant and space-variant diffractive optical interconnection devices, employing VLSI-based patterning and dry etching technology. We also discuss the realization of real-time reconfigurable optical interconnects using phase-modulating spatial light modulators.

Current planar integration technologies utilize one or two dimensional light propagation in waveguides. Technologies to achieve microintegration of classical optical components in three dimensions are available.

Optical image crossbar switch (OPIX), based on a multiple imaging system, is proposed and demonstrated. OPIX is reconfigurable and nonblocking switch for two-dimensional images. As OPIX uses geometrical optics, it has a large space bandwidth. The proposed OPIX is incorporated into a newly proposed optical parallel processing system and initial tests are encouraging.

The permutation of switches is applied to generate (1) butterfly, (2) crossover and (3) global/local interconnection patterns. Both classes, the shuffle-based interconnections (1) and (2) and the global/local interconnections (3) are analyzed and compared.

We propose a new planar device for reading light supply on a spatial light modulator. The new device, reading light supplier (RLS), consisting of a fiber plate and a redirector was found to be a suitable device for stacked optical computing systems. A fiber plate has a normal light propagation mode and a spreading light propagation mode, both of which are efficiently utilized in a RLS. The redirector placed on a surface of the fiber plate redirects the propagation direction of light in the fiber plate. Elementary experiments were carried out with RLSs, and we can successfully read out 2-dimensional images.

In this work, we investigate various approaches to reduce the overall complexity of optical MINs. We focus on a volume Omega network which uses 2-D perfect shuffles for the fixed interconnection at each stage. In the 1-D perfect shuffle, the N channels (assuming N is even) are divided into two halves, which are then interleaved perfectly. In the 2-D separable shuffle, the rows and the columns are shuffled independently by a sequence of 1-D perfect shuffles. The 2-D folded shuffle is equivalent to performing a 1-D perfect shuffle on a 1-D array folded onto a 2-D plane. The rows and the columns of the input are obtained by folding (raster scanning) an n² X 1 1-D input array onto an n X n 2-D input. A 1-D perfect shuffled version of the input can be obtained after unfolding the output plane of a properly shuffled 2-D output. This paper describes the architectural design of free space volume Omega networks using simplified hardware. Optics are used to perform the 2-D shuffles, and optoelectronic switch element arrays are used to perform the dynamic switching. Our main observation in this paper is that the regularity of the fixed link patterns at the stages allows us to use a single set of optics to perform 2-D shuffles for all the stages. We discuss two techniques for spatially multiplexing the channels so that all stages of shuffles can be performed by a single set of optics.

Hologram amplification was found for the first time in the crystals of lithium niobate LiNbO₃. This phenomenon has been observed in several kinds of recording media. The detailed phenomenological description of that effect is given. The photopolymer materials (PPM) are organic light-sensitive media. The light-induced radical polymerization in these materials causes refractive index increases. Increased diffraction efficiency of holograms in PPM due to one-beam postexposure was observed for the first time in the first half of the 1970s. The hologram amplification in the liquid photopolymer FPK-488 composition has been recently investigated in details. The postpolymerization amplification (PPA) in thermal reactions has been noted. The results of experiments for the coherent dynamic self-amplification (DSA) in that material are presented. The amplification mechanism has also been analyzed in these articles. The Du Pont de Nemours photopolymer materials allow thermal amplification of reflection holograms. Samples are heated to the temperature of 80 degree(s) - 125 degree(s)C after the recording and during 180 min. treatment the refractive index increases from 0.007 to 0.015 - 0.022.

A theoretical and experimental investigation on the far-field diffraction patterns of a dot array with Gaussian random fluctuations has been done quantitatively by means of a generalized fluctuation parameter. It has been found that the exposure condition strongly influences the nature of the observed photographic pattern and hence the observation of the light-depletion phenomenon first reported by Stark and later, analyzed by Martin and Aime.

Microlenses in polymethyl methacrylate can be fabricated by irradiation with a high energy proton beam and diffusion of monomer vapor in the irradiated domains. The relative aperture can be as high as one. The fabrication process, results and applications are discussed.

Holographic disk based photonic neural network with outer product implementation is described. Some aspects of two-dimensional fourier-holograms registration on the photothermoplastic disk under infrared laser thermodevelopment are considered. Developed approach allows to supply the high information recording density (approximately 10⁵ bits/mm²) for the holographic disk, the high quality of IWM fourier-holograms registration ((eta) equals 5 - 7% and contrast ratio approximately 70:1) in addition to the possibility of image reconstruction from the hologram identical to the original image without aberrations.

We demonstrate self-aligned connections at 633 nm between single mode fibers with two types of mutually pumped phase conjugate mirrors, made with photorefractive BTO crystals. The first type of device is the two-facing double phase conjugate mirror and the second is a double phase conjugate mirror, splitted into two zones. This second configuration provides higher efficiency and shorter response time (about 15 s). Up to 70% of the conjugated beam intensity may be coupled into the output fiber, while the total efficiency may reach 5%, mostly limited by crystal absorption. The influence of the period and orientation of the grating is experimentally studied. Reconfiguration capability is shown with reconfiguration time of a few tens of seconds. Efficient injection from a multimode fiber into a single mode one is also demonstrated. Possible applications of these self-aligned connections to free-space switching systems are suggested.

An optoelectronic implementation of the Hopfield model is presented. 2-D PDA is used for vector-matrix multiplication of binary optical signals. Multi-level interconnection matrix is realized by accumulating signals on PDA during several cycles.

A optical perfect shuffle interconnection and array of exchange boxes are given. The optical perfect shuffle network is implemented using repeated stages of PS together with arrays of exchange boxes.

Continuous surface-relief phase gratings for two-dimensional (2-D) array generation have been designed and fabricated by laser-beam writing lithography. Separable and non-separable solutions for the design of 2-D fan-out elements are compared. A 9 X 9 fan-out element has been fabricated in photoresist by laser-beam writing. A diffraction efficiency of 94% and an uniformity better than +/- 8% over the whole array were achieved.

The way to increase signal intensity from photon echo memory cell by interference of several stimulated photon echoes is analyzed. We demonstrate that small fluctuations of optical delay time neglected thus far can dramatically affect this interference. New interpretation of an experimental result is obtained on this basis.

An optical system is presented for dynamically reconfigurable local interconnects, based on liquid crystal retarders and birefringent crystals. Quantitative measurements illustrate actual performances. Applications are discussed.

A design and a fabrication of multiple beam generators using zone plate array are reported, considering intensity distribution of a Gaussian beam. A multiple beam generator is composed of 10 X 10 circular zone plates with 5 kinds of zone radius ranging from 0.42 mm to 0.85 mm, considering intensity distribution of a Gaussian beam. The focusing characteristics are examined at 0.78 micrometers in wavelength of LD under normal incidence condition. The focusing intensity nonuniformity was corrected such that 10 multiple focusing spots with relative intensity fluctuation less than 20% were achieved. Feasibility of application to high speed opto-electronic integrated circuit is discussed.

Deep etch lithography of PMMA by the LIGA-technique or H⁺ lithography is a powerful tool for the fabrication of very high refractive microprisms. We have realized a microintegrated optical imaging system using microprisms fabricated by the LIGA-process and microlenses with a diameter of 250 micrometers . Thus the size of the imaging system is less than 2 mm⁺.

Diffusion elements are fabricated by locally manipulating the ion composition of a glass substrate. For our experiments we use the silver sodium ion exchange technology. The effort of our work concentrates on the generation of arbitrary index distributions. Therefore we characterized two different types of special glasses--one of them showing linear, the other nonlinear diffusion response. To be able to optimize masks for the generation of arbitrary phase profiles we simulated the diffusion process numerically.

Direct image transmission through a multi-mode fiber is demonstrated using an optically addressed spatial light modulator as a phase conjugation mirror with an image-input function.

We consider the problem of identifying each of multiple objects in a scene with object distortions and background clutter present. Attention is given to the role for neural nets (NNs) and optics and the type of NN used. A hierarchical/inference system is used with correlation neural nets used for low levels and new NNs with higher-order decision surfaces used for classification.

A polychromatic optical neuro-computer for color pattern recognition with cascaded liquid crystal televisions (LCTV) is presented. Extension of this neural net to multichannel operation is proposed and preliminary experimental results are presented.

We proposed a new renewal method of the weights to realize a back propagation learning algorithm with a large number of iteration times in an optical neural network. The main purpose of this method is to use the limited dynamic range of an optical device widely. Since some weights displaying on the device are quickly saturated in the conventional method, the learning is not completed in many cases. The new technique can suppress saturation of the weights so that learning is completed. The computer simulation and the optical experiment are presented.

A complex phase-conjugate neural network model with a Hopfield-like energy function has been proposed, and a physical interpretation is given to the model and its dynamics. The results of experiments and computer simulations are presented that demonstrate the behaviors of the complex neural fields predicted by the theory.

Different rules for weighting input vectors in high-order associative memory models are considered. Numerical simulation results are given showing the advantage of the pattern weighting in recognizing high-correlated patterns. An optoelectronic implementation of the inner-product associative memory with weighting method is presented.

One compact optical architecture for implementation of second-order neural network, which get rid of some redundant terms of the input vectors and the interconnection matrix without reducing its associative memory, is put forward.

On the base of probability density functional (PDF) approach a self-learning neural-network is synthesized for the restoration of an arbitrary atmospherically distorted coherent images. Learning capabilities and potential accuracies of the synthesized algorithm are studied. Results of numerical simulation of the method are discussed.

A flexible system has been set up for image processing, with prototype liquid crystal input, BGO filter and photothermoplastic programmable memory. First experiments utilize fixed references for invariant pattern recognition. The final system will implement a neural network classifier.

We review recent work on rotation invariant pattern recognition. We briefly review results involving Circular harmonic filters, then describe some of our more recent research in this area. A pattern classification system that achieves rotation invariance by means of an optical rectangular-to-polar coordinate transformation is described, as well as a system using circular sampling. Experimental results using optical systems are presented for all those methods.

A joint transform correlator with feedbacks from the correlation outputs to the input intensity of the reference images works as a neurocomputing system. The characteristics change catastrophically with the parameters of the feedback transfer function.

Optical cellular logic and extended-interconnect cellular logic processor architectures are described. Simulations, tolerance studies and circuit construction are used to advance both device technology and architectural thinking. Benchmarking against electronic distributed array processors leads to an optical non-locally interconnected, off-chip scheme with local electronic connections between smart pixels, unless ultra-low switching energy all-optical components can be fabricated. Fundamental limits to optical bistability do not rule out the latter possibility.

Since optical communication is preferable for establishing connections exceeding a certain critical length, for large system sizes the beneficial use of normally conducting wires for the shortest connections becomes an edge effect and can be ignored. This suggests that the performance and cost of an all optical computer might not be much inferior to an optimal hybrid alternative. We argue that for applications for which high bit repetition rates are useful despite large propagation delays, it might make sense to contemplate the construction of an optical digital computer.

We demonstrate and characterize the use of an optically addressed ferroelectric liquid crystal (FLC) spatial light modulator (SLM) as a spatial filter. The photosensor associated with the liquid crystal is a PIN photodiode made of hydrogenated amorphous silicon (a-Si:H). Both the amplitude and the phase of the reflection coefficient are observed to be modulated by the write beam. The filter function can be reconfigured at a submillisecond rate with incoherent illumination.

Implementation, device requirement, and processing capability of a parallel optical computing system based on image logic algebra (ILA) is investigated. The proposed optical implementation requires three types of optical devices. The processing capability of the ILA based system is discussed in terms of numerical processing power.

The incoherent optical correlator was extended to perform a rotation invariant pattern recognition by using a set of two real-positive filters presenting a single order circular harmonic function.

Novel optoelectronic architecture of an arithmetic logic unit is proposed using self-routing optical interconnection of laser diodes. Basic experimental results of the unit using twin- striped beam-scanning laser diodes are presented.

An experimental optical digital parallel processor has been constructed. 256 channels are synchronously clocked round a three-gate loop containing a programmable logic gate, a fan- out and threshold stage, and a memory gate.

The problem of working out and research of the high efficient special-purpose optical- electronic computer system for invariant pattern recognition are being considered. The computing processes organization method has been suggested, being the basis for the development and technical realization of the reconfigurable multiprocessor optical-electronic computer system. The evaluation of the system speed has been given.

Theoretical investigation of four- and three-level fiber amplifiers based on Nd³⁺- and Er³⁺- doped silica fibers has been carried out. Analytical expressions for the gain factor and the power of the output signal saturating the gain factor are obtained. We have showed that the corresponding choice of the refractive index profile allows to increase essentially the power of signal saturating the gain.

All-optical gates based on nonlinear interaction between two oscillations in similar nonlinear resonators or two waves propagated down two similar nonlinear waveguides are considered.

Mathematical description and algebraic interpretation of recursive algorithm in optical binary pattern processing are carried out. The recursive optoelectronic processor configuration is described.

The main parameters of an image crossbar switch implemented by means of polarizing beam slitter cubes, switchable half-wave plates and objectives are analyzed. It is shown that these crossbar switches can have a data capacity of 1 Pbit/s and a reconfiguration time to 10 ns.

Realization of a superfast multiprocessor all optical computer based on some broadband optical logical gates with large delays is considered. It is shown that total performance is determined by duration of the light signals and doesn't depend on the value of delays in transmission signals between gates in first approximation for the great classes of solved problems.

We propose a method for analyzing interferograms which relies on optical spatial filtering. This method permits performing the necessary mathematical operations on the original hologram by means of optical analog calculations. As a result, one can quickly obtain information about the parameters under study in the form of two-dimensional intensity distribution.

We discuss our cascaded correlator-based optical numeric processor and its projected performance (our goal is a numeric processor and not a general-purpose optical processor). We use symbolic substitution (for parallelism on long words and arrays of words), the modified signed-digit number representation (for speed, i.e. reduced carries), and a new encoding and substitution architecture to improve performance.

Novel architecture of an optoelectronic digital computer for cellular matrix multiplication based on the binary sign-bit number system is proposed. Such a computer comprises dynamic memories based on circulation fiber-optic delay lines, which ensures loading of processor elements up to 90%. Owing to the use of cellular logic algorithm in the optoelectronic computer, it is possible to perform multiplication on matrices of any dimensions employing only one base element. By expanding functional capabilities of the elementary processor, the proposed specialized computer can be transformed into a multipurpose optoelectronic computer.

On the basis of the investigations of the interference field spatial structure methods of optical analog calculations are proposed: differentiation of the functions of three variables, calculation of gradients, linear integrals of vector functions, and curvilinear integrals.

We present the OEDIPE project (Opto-Electronic DIgital Processeurs Elementaires). We explain the motivations for massively parallel opto-electronic computer and describe the architecture of the computer, we point out the optical interconnections based on fan-in and fan-out functions and opto-electronic technology suitable for the project.

Spectral properties of the dynamic Weigert-effect in the media based on bacteriorhodohpsin have been studied both experimentally and theoretically and possibility of the real time proceeding of color images have been shown.

An optical learning network using photorefractive holograms with grating amplitude detection is presented and demonstrated. The Stoke's principle for light is used to realize both bipolar weight changes and bipolar activities for perceptron learning. The bipolar detection is achieved by using a double Mach-Zehnder interferometer and an electronic subtractor at the output port. Experimental results show that a zero or near to zero threshold value, the network can be trained to dichotomize the input patterns.

Different linear and non linear edge extraction methods applied as preprocessing to the scene to the target or to both are studied in pattern recognition based on correlation. Results for binary and grey level objects are given.

An experimental and numerical study of new optical elements is outlined for the realization of differentiation as well as Hilbert and Fourier integral transforms of Fresnel images. The elements mentioned in the form of amplitude and phase gratings consist of two parts of different partition width-to-their disposition period ratios. A lensless mapping of two- dimensional Hilbert transform has been found experimentally by means of a grating grid consisting of two orthogonal gratings considered. The possibility is found of beam differentiation by means of a periodic grating permitting to eliminate the effect of the diffraction structure center shift in respect to the beam axis on the image formation. A substantial doubled image dependence on the input beam wavefront curvature has been revealed. The geometry of a two-dimensional spacial filter for optical beam switching is proposed.

The management of current and future very large databases necessitates the development of secondary storage with massive capacity and transfer rates as well as highly parallel processing techniques. Parallel read-out optical disks can satisfy both memory requirements and, thus, constitute a suitable candidate for database storage. In this paper, we describe the architecture and functionality of an optical database filter based on parallel optical disks and capable of performing several relational database operations. As the disk rotates, records are retrieved in parallel and compared to search arguments loaded in spatial light modulators. Only data that satisfy a query are allowed to pass through a photodetector to an electronic buffer for further processing. Initial performance analysis has shown that the system is capable of processing data at a rate exceeding one million tuples per second.

A new synthesis method for implementing arbitrary n-input Boolean functions using optical thresholding logic schemes is proposed. The logic block consists of an exponentially weighting step and a multi-thresholding step. Programmability and cascadibility are available.

We demonstrated a new method to perform matrix-matrix multiplication by using anisotropic self-diffraction in BaTiO₃. The digitalized output is suitable for large scale communication and interconnection. The Bragg conditions, dynamics of output intensity and operation speed of the architecture are analyzed.

Based on the characteristics of the symmetric self-electro-optic effect device (S-SEED), a logical model of the device is developed. This model can describe all the functions which an S-SEED can perform as a latch or as a logical OR, AND, NOR, or NAND gate according to different optical preset and input signals. The model can be used for system simulations. It forms a basis for the design of digital optical systems based on S-SEEDs.

We demonstrate the calculation of the auto-correlation matrix of a vector by optical processing, using an array of asymmetric Fabry-Perot (AFPM) devices. This is a basic requirement for a second order neural net work, and so opens up the potential of using AFPM device arrays in second order optical neural networks.

A new method is presented for shifting optical information signals in the plane of optical bistable array in a direction normal to the direction of incident light beams. All-optical version of shift register and CCD memory device is experimentally demonstrated using semiconductor nonlinear interference filter. The possibility of creating more complicated optically controlled planar circuits and networks is shown.

A complete parallel version of hybrid optical parallel array logic system (H-OPALS 16²) is constructed, which can process 16 X 16 pixels simultaneously at up to 19 kframs/sec. In the system, all pixels are processed and transferred in parallel. Experimental results, including parallel numerical processing and image processing, show capability and flexibility of the system.

A cellular logic image processor was designed, constructed and successfully operated by interconnecting two arrays of symmetric self electro-optic effect devices (S-SEED). This paper outlines some of the design issues associated with the implementation of a free-space digital optical system.

The device uses pattern recognition techniques and diffraction properties of particles. A light modulator made of a photoconductive crystal associated with liquid crystals, measures particle size in real time.

A fully functional optical adder based on Systolic Arrays and Symbolic Substitution (SSL) is demonstrated. It is implemented in a modular, miniaturizable and micro-integrable optical setup. The optical system is designed to fit the requirements of incoherent optical processing with optoelectronic threshold amplifier arrays. Parallel application of four SSL rules is performed in a single, light efficient optical path.

Methods for designing the architecture of optical multiplanar array processor are developed. The underlying abstract model (Parallel Substitution Algorithm and Parallel microprogramming) and associated theory of parallel distributed computations are briefly described. A general method for determining cell logic and structure of the array which implements a fine-grained parallel algorithm into an electrooptic processing unit is proposed. Formal tools for transforming 2D-cellular algorithm into a 3D-representation is given in a short form. Method proposed provide formal ways of creating facilities for computer aided design of algorithmically oriented electrooptic architecture.

The design criteria of an all-photonic error control system are presented. It is implemented by cascading a number of iterative cells. The performance is evaluated and optimized by simulating its behavior with the Beam Propagation Method.

Recent construction of digital optical processors has highlighted the need for accurate control, measurement and modelling of system and device operating characteristics. Study of the single channel O-CLIP allows the optimization of speed, algorithms, and architecture design.

A systems is described which finds solutions to the 6-city TSP using a Kohonen-type network. The system shows robustness with regard to the light intensity fluctuations and weight discretization which have been simulated. Scalability to larger size problems appears straightforward.

Picture type analog-to-digital converters (PT ADC) creation necessity for the parallel digital optoelectronic processors are shown. Various approaches to the PT ADC construction are considered. Achievable parameters of PT ADC under using of various kinds of elementary base are estimated. 2D-element geometry noise influence on the PT ADC accuracy are investigated with the help of computer modeling.

Additive noise in optical imaging systems puts a fundamental restriction on the reliability of object localization and recognition in pictures. This is analyzed in the paper as a trade-off between the signal-to-noise and picture size-to-object size ratios or quantity of objects to be discriminated.

An optoelectronic multistage compact interconnection network using polarizing beam splitters, polarizing switch arrays, lenslet array and free-space optics is proposed. It is capable of realizing a interconnection of a large number of data pattern images with the total data throughput up to 1 Pbit/s and a reconfiguration time up to 10 ns.

Nowadays massively parallel computer systems consisted of many thousands of processor elements (PE) are reality. Their main component is so called communications network (CN) that provides for data exchanging between PEs. CN represents most of the cost of the computer, most of the power dissipation, most of the wiring etc. CN consists of communications processors (CP) or routers. Each PE has its own CP. CPs are connected by transmission lines in some regular way determining CN topology. To use the advantages of broadband optical transmission lines most efficiently it makes sense to take an approach when an optical computer consists of a number of identical virtual PEs, i.d. presents massively parallel computer1. One actual optical PE implements N virtual PEs , where N=T/T, T is the delay time in propagating signals between all-optical gates, Tis the period of their switching. N is usually equal to iO-iO. Time interval T is divided into N identical time-slots. Each of the virtual PEs operates in its own time-slot. The number of PEs coincides with the number of timeslots. Thus the data transmission between different PEs reduces to the data redistribution between time-slots. We have drawn the conclusion that N virtual CPs can be obtained on the basis of one actual CP being consisted of the above-mentioned optical gates. CNs of various topologies making possible a necessary trade-off of the number of gates against the transmission speed can be easily designed. The total number of transmission lines in the optical CN is N times less than it is in the similar electronic CN. The most simple an optical CN of 2D grid type consisted of n columns and N/n rows of PEs has been considered before1 . There is also a universal time-slot interchanger that is the time domain equivalent of a multiple input, multiple output spatial interconnection network2. There are 2n-1 optoelectronic switches connected in series for universal time-slot permutation of serial 1-bit array of size N=21 time-slots. Two basic features of the interchanger ought to note. First, the optoelectronic switches are controlled by electrical signals generated by special electronic generator. This factor limits the decrease of value of time-slots and prevents the taking full advantage of very high potential of broadband of optical channels. The electronic gates used in the generators could be used instead of the optoelectronic switches also. Delay lines could be replaced by electronic shift registers. Second, there is no self-routing. The permutation is defined by some external electronic generator. These features limit the possibilities of optical multiprocessor computer.

Adaptive optical systems can be used to clean up atmospherically disturbed beams or to predistort a clean beam so that it phase compensates the atmospheric disturbance in passing back through the atmosphere (1-4). Such systems need to deal with the whole wavefront to smooth it, so global measurements must be processed to obtain that information. This mitigates against local, pixel-by-pixel independent operation of the active phase controller (SLM). Thus, a tension arises between the need for a global fit and the speed and convenience of purely local operation. A description of the entire front of the wave is necessary because no local measurement such as made by a shearing interferometer or the Hartmann test, can separate a global trend from a local fluctuation. These samples of derivatives must be integrated into a global pattern. On the other hand, global calculations take time. A purely local, adaptive optical system would be simpler and faster. We will show how to use global self—referencing to allow local adaptive optics to be used in a "pure optical" system - no conventional, discrete component optics will be needed, in principle