New correlation filters for object detection are considered. Detection is the first step in general scene analysis. It requires locating the positions of all objects (regions of interest) in a scene. This involves recognizing objects in multiple classes with object distortion and contrast differences in the presence of clutter. To solve this problem, we use basis function descriptions of objects using new Gabor wavelet (CW) filter functions with new correlation plane optimization functions and employ degrees of freedom (DOFs) to satisfy the optimization functions in a controlled manner. We refer to these as DOF filters and to our specific versions as GW-DOF filters.

A novel optical wavelet processor is introduced. Generic wavelet functions, consisting of both amplitude and phase information can be optically synthesized using two cascaded Liquid Crystal Television Spatial Light Modulators. An innovative ternary-valued shape discriminant wavelet filter and its optical implementation is also introduced. Experimental demonstrations of multitarget classification using both a 2D Morlet wavelet filter and a ternary wavelet filter are also reported. The experimental results show that the wavelet filters are superior in discrimination than that of the conventional correlation technique.

A filter function is derived for input signals containing a target which is spatially disjoint (that is non-overlapping) with the input scene noise. The optimization metric is the ratio of the square of the expected value of the correlation peak to the expected value of the output signal energy. In this model, the effects of the non-whiteness of the scene noise, the non-stationarity of the scene noise due to the limited size of the input scene, the non-overlapping of the target and the scene noise, and the unknown variations of the target illumination are considered. We show that, for the non-overlapping target and scene noise, the target window and the scene noise window strongly influence the optimum filter function.

This paper presents optical laboratory data on a number of new optical filter systems for: rank-order morphological filtering, morphological ternary phase amplitude filters, morphological hit-miss detection filters, Gabor detection filters and distortion-invariant detection filters. All filters and processing are performed on the same optical correlator architecture. This provides a general purpose multi-functional optical image processor for general scene analysis, capable of low-level vision, detection, and image enhancement operations.

Zero and fifteen phase-state filters designed for training sets of zero mean imagery are discussed. These filters are demonstrated to be highly invariant with respect to targets which are simultaneously translated with respect to a background, to changes in the background, and to additive zero mean noise.

Phase-only filters are attractive designs because they provide good noise resistance and fairly narrow peaks together with optimum light efficiency. Since they contain only phase information they are easier to implement in current spatial light modulators. However, their extension to the synthetic discriminant function (SDF) formalism is troublesome owing to the nonlinear constraints involved. We developed a new procedure for computing phase-only SDF filters which is intended to overcome some of the pitfalls of previous methods.

Typical coding domains, including phase-only, binary phase-only, amplitude-only, spiral and ternary codings, are compared from the point of view of their filtering performances. This comparison is based on a multicriteria optimization method recently introduced for the determination of optimal trade-off filters for implementation on arbitrary spatial light modulators in the Fourier plane of an optical correlator. Constrained filter design examples are presented for illustration of the method.

A correlator performs better if the input and filter are made compatible with each other. To date, most quantitative work has been spent on improving filters. We give general rules for mutually optimizing input image conversions and filter SLM control. The rules take into account several realistic system considerations, including phase and amplitude coupling of input and filter modulators, inherent system noise, and a wide variety of quadratic ratio metrics.

We show laboratory results showing coupled-filter improvements in certain correlator performance metrics: signal to noise ratio, peak correlation energy, and peak to total energy. We had previously shown improvements in correlation intensity. The filters were built using JSC computer code for minimum Euclidean distance optimal filters.

Epson LCTV has been used as a phase only filter in a correlation set up for pattern recognition. The phase modulation of Epson has been measured using a Mach-Zehnder interferometer. We implement the phase only filter on the liquid crystal using different levels of quantization in the phase. We compare the impulse response and the correlation results obtained with this filter with different levels of quantization. Finally we compare the results with those obtained using high resolution techniques as lithography to produce computer generated holograms.

When real time optical pattern recognition is carried out by means of spatial light modulators (SLMs), a practical situation may arise in which the size of the Fourier transform and the intensity of the light source are fixed. Then, provided the input scene and the size of the SLM allow it, an interesting possibility for improving the overall intensity in the Fourier plane is the replication of the scene. This procedure may be used in two different ways: regular along a lattice structure or non regular or random replication. In this work we analyze the usefulness of these methods, both to obtain diffraction images or cross-correlation between images in a joint transform correlator.

An implementation of a phase extractor Vander-Lugt correlator, which operates with a single spatial light modulator is suggested. Optical phase-retrieval manipulation, based on the symmetrization of the input scenes is proposed. As a consequence, the number of digital manipulations of the information is substantially reduced and spatial correlation filters can be used. Theoretical analysis and some simulated results are provided. The study also includes the results when the scenes are immersed into background or zero mean additive noise.

An all-optical video alarm system was constructed and tested to demonstrate unsupervised learning. Previously we showed experimental results in which Hebbian learning was used in a supervised learning system for optical heteroassociative memory. We show the relation between Perceptron learning and supervised Hebbian learning and then show how the latter may be evolved into the unsupervised Hebbian learning used in this paper. The concept is first explained using an optical perceptron like set up with a delay and a spatial light modulator made of electron trapping material. The delay is accomplished using two SLRs. Two cycles are used in a folded system so that the same image intensifier/liquid crystal light valve combination may be used for reading both SLRs. Experimental results show the correct functioning of the system.

We describe and illustrate a hybrid statistical-syntactic pattern recognition system which seems to offer the simplicity of statistical pattern recognition with the power of syntactic pattern recognition.

The principle and experimental results of a novel spatial bifurcating optical associative retrieval and pattern recognition technique are presented. A single 2D input image is associated with self-amplification with two output images through holographic diffraction in a photorefractive crystal. In an optical experimental setup utilizing a photorefractive crystal BaTiO₃, interesting optical edge enhancement, pattern recognition and auto-associative retrieval phenomena have been demonstrated.

We present a large-scale holographic associative memory system of binary images, with which an electronic digital computer is incorporated to realize a high performance and to combine a logical processing using external constraints. In the system, associative process is performed in two steps, correlation and image synthesis steps. In the correlation step we execute a simple nonlinear operation, and in the image synthesis step we introduce an external input based on the prior knowledge about the expected output. This external input lets us control the output of the system. In the optical implementation we use spatially multiplexed Fourier transformed holograms. Experimental demonstrations are given to illustrate the performance of the system and the effectiveness of the external input.

1D hologram technology for information processing is considered. A high-quality neuroprocessor with serial-parallel processing is presented. It is shown that a high-capacity neural network having up to 10¹⁰ interconnections may be implemented with a small number of real neurons. The developed 1D holographic disk memory allows the implementation of 3000 different neural networks with the number of each structure being equal to 1000. For this case the processing time is presently equal to 70 ms. A holographic associative memory with 1D pattern recording using a photothermoplastic carrier is investigated. The results of associative information retrieval according to presented pattern fragments are given. The array selection error probabilities versus the number of bits in the key array and the number of recording/erasure cycles were measured.

Noise performance of an optical correlator for credit card verification system is investigated. Various types of multiplicative input noise, including amplitude noise, phase noise and amplitude-phase noise, are considered.

With recent advances in the state-of-the-art in spatial light modulators, the optical joint transform correlator and binary joint transform correlator are becoming practical signal processing systems. However, the performance of these devices is severely limited because of the dominance of the reference and scene autocorrelation signals in the output plane. Two major problems caused by this are false correlations between multiple targets and low correlation peak amplitudes. Two recently proposed techniques which reduced or eliminate the autocorrelation signals are Fourier plane windowing and time modulation (or frame subtraction). This paper shows how these techniques can be used separately or together to improve binary joint transform correlator performance. In simulation, the two techniques provided peak to noise improvements of 2.5 and 5.5 dB over binarization based on using a uniform threshold. The improvement gained by combining the techniques was 6.6 dB.

It is known that a transition from object to spatial-frequency domain with the aid of the Fourier transform may be a suitable solution for the problem of real-time pattern recognition of objects whose features cannot be determined exactly. The paper discusses the method of incoherent space frequency analysis with the aid of liquid-crystal light modulators. The hardware based on this technique can be simple and economical, and it overcomes a number of disadvantages of coherent-optical set-up. The parameters of LC modulators are discussed. These may be used to derive operational conditions, processing spreads, etc. of the optical processors involved. Selected examples illustrate suitable applications.

We present a technique for the optical implementation of a color image correlation using a liquid crystal spatial light modulator (LC-SLM). A color image composed of 3D vectors of RGB primaries, is mapped onto a 2D plane to extract chromatic information from the image. Then, the 2D vector image is defined as a complex function, which is displayed on the LC- SLM and employed in a complex correlation system. Results from computer simulation and an experiment are also demonstrated. Optical parallel recognition processor for color image recognition is realized by the coherent optical system of single wavelength.

Within this paper we discuss the precision of a fixed-point discrete numeric vector-matrix processor. The concept of precision relates to the quantity of numerically exact information in the processor's calculated result. This analysis is based on a signal space formulation which allows for the determination of precision from the uncertainty in the output signal space. Characteristics of the ideal signal space are explored and then the implications of simple nonideal temporal and spatial effects are considered. The resulting precision limitations are discussed in terms of numerical roundoff and significance errors.

Large, flexible space platforms, such as the proposed Space Station Freedom and Earth Observing System, may experience vibrations caused by astronauts, motorized instruments, or cooling systems which disturb sensitive pointing operations elsewhere on the platform. Linear control systems can be used to dampen these vibrations. In this paper, we discuss the application of optically implemented linear control algorithms to a truss structure model as NASA's Ames Research Center. We describe the system architecture and discuss the effect of reduced accuracy in linear control systems due to analog computation.

This paper describes the design and fabrication of a high performance optical vector-matrix coprocessor for optical computing research applications. The optical vector-matrix coprocessor is configured to multiply an 8-element vector by an 8 X 8 matrix with a throughput rate of 1 MHz--effectively achieving a processing rate of over 100 Mops. The Vector-Matrix Coprocessor interfaces to an industry standard Personal Computer with a single card and is controlled by software written and compiled in the ANSI C language. All data input and output to the coprocessor are in 8 bit digital words. An 8 to 12 bit look up table is provided for each input channel to provide real time linearization of analog optical data representing input values through the optical system. The optical signals representing calculation values are detected and received by a switched capacitor integrating filter to reduce detection bandwidth and reject broadband noise.

In this paper, we present the first analysis of the accuracy of both addition and multiplication on a multichannel optical matrix processor. We performed our analysis on an analog optical vector-matrix processor capable of performing 120 M operations per second with an 8-element input vector, 8 X 8 matrix and 8-element output. We examine the accuracy with which numbers are represented for both the acoustooptic vector input and the liquid crystal matrix input, the accuracy of multiplication on a single channel, the accuracy of addition on two channels as well as the accuracy of multiplication and addition on two channels. From this analysis, we can conclude that while vector input numbers are represented with 7.7 bit accuracy, the results of two-channel multiplication and addition are only accurate to within 1.3% or 6.3 bits.

This paper presents an algorithm for constructing a ternary phase-amplitude synthetic discriminant function filter tree for classifying an unknown image. The tree can classify an image in 2(DOT)Log₂(N) filter applications, where N is the number of training set images used to construct the tree. An example tree is constructed to demonstrate the concept. The example tree contains 10 M60-A1 images and 10 T62 images from 0 degree(s) - 90 degree(s) orientation. The resulting tree is applied to 162 test images of M60 and T62 tanks that were not included in the training set. The tree was 100% accurate in classifying each test image as the correct type of tank, and was 94% accurate in identifying the training set image that each test image was closest to in orientation.

In this paper, a new approach for uniquely selecting the optimum reference patterns of a matched spatial filter (MSF) is described. By the reference patterns we can synthesize the optimum MSF despite the target patterns are changed from letters to numerals, for example. In the experiment, two types of the reference patterns are used. One of the types is called primitive patterns which contain a vertical bar, horizontal bar and circle patterns. They are all orthogonal to each other in the view of image processing which means that three patterns are not similar shape at all. The other type patterns are given by subtraction of a few target patterns which the primitive reference patterns can not recognize. Using these reference patterns, an MSF was synthesized which could fairly recognize the 26 English alphabet letters.

We present the eigenpolarization states of a commercially available liquid crystal television display and show that phase-only modulation can be achieved over a large dynamic range of video voltages for several bias voltage settings if the eigenpolarization states are used. A set of operating curves using these polarization states are given for the device.

Ancient cuneiform inscriptions are remains of historic and cultural value. The characterization of cuneiform signs by means of feature enhancement, feature extraction, detection, and identification is therefore of special interest. The use of methods minimizing the in-class objects sensitivity while maximizing the discriminability for out-of-class objects is indispensable due to the inherent sensitivity of optical pattern recognition. To analyze cuneiform inscriptions a multifunctional experimental device was advanced enabling correlation experiments to be performed including an automatic prefiltering procedure. An original technique for coherent optical averaging is introduced. The in-class and out-of-class correlation sensitivities are compared for different types of matched spatial filters including classical single sign filter, optically averaged filter, digitally averaged filter, and multiplexed filter.

A novel opto-digital pattern recognition method which has the shift, rotation, and scale invariant properties is proposed for recognizing 2D images having straight lines. The proposed method is composed of three stages. In the first stage, the line features of the image are extracted. The second stage imposes the shift, rotation, and scale invariant properties on the extracted features through the proposed operations. In the last stage, an artificial feed forward neural network is trained with the extracted features. To evaluate the proposed algorithm, nine different binary edge enhanced images are utilized as the experimental patterns. The successful recognition results through the computer simulation and the opto-digital experiment are presented.

In this paper, 2D optical code division multiple access (OCDMA) networks based on the costas array are compared with the 1D OCDMA networks through the computer simulation and some simulation results are also provided. It shows that a 2D OCDMA network using costas array allows shorter bit times, for a given laser pulse width, and 2D codes as a costas array result in a reduction of autocorrelation sidelobes and crosscorrelation peaks, and this network has lower losses compared to the 1D OCDMA network.

An experimental investigation of the scale and rotation invariant joint transform correlator (SRIJTC) is presented. The SRIJTC is a three-step algorithm, one more than the conventional JTC, and where in the first step we perform a logarithm-polar coordinate transformation to achieve scale and rotation invariance. The mask for the logarithm-polar coordinate transformation is obtained with a Lee's binary interferogram computer generated hologram with a mapping of 4(pi) . The product between the coordinate transformation mask and the input signal is made digitally and displayed on a liquid crystal television to perform the logarithm-polar coordinate transformation. The optical setup used to obtain coordinate transformation is the same that perform the JTC leading to a three-step JTC invariant to scale and rotation.

A multichannel joint transform correlator, using a Dammann grating as a beam-splitter to split one incident beam up into 2D array of equal intensity beams to form multichannel for correlation, is presented and investigated. Mathematical analysis and optical experimental results are presented.

Image Processing is widely regarded as a successful application of electronic technology. However, there is a wide gap between commercially successful image processing implementations and system design that have been proposed by the research community. Of particular interest are image processing architectures that use large number of densely interconnected processors to compute and transmit image date in parallel 2-D format. The interconnect density and packaging constraints of electronic packaging technology preclude efficient implementation of such architectures. In contrast, optoelectronic integrated circuits combined with surface-normal optical interconnects offer the promise of a technology platform that is inherently better suited for the implementation of interconnect intensive image processing applications.

We present the Stretch switch; a new class of self-routing multistage interconnection networks that provides a continuous performance-cost tradeoff between two of its degenerate forms: the Knockout switch and the Tandem banyan network. Stretch networks utilize simple destination tag routing and can be designed to achieve low delay and arbitrarily low blocking probabilities for random, permutation, and non-uniform traffic without using internal buffers in the switches. These qualities make them ideally suited for both fast packet (ATM) switching and multiprocessor architectures, and facilitate efficient VLSI and photonic implementations.

We extend a serial communication technique, Partial Response (PR) precoding, to two and three dimensions to increase the capacity and signal separation in parallel readout optical memory systems. We also develop expressions for optimal 2D equalizers that improve signal separation by reshaping reconstructed pixels. We simulate and experimentally verify the performance of various forms of PR precoding in a coherent imaging readout system using 2D arrays of binary phase storage pixels. However, the techniques presented here may be applied to other readout systems and other storage values. With an equalizer, PR precoding yields a factor of two improvements in worst case signal separation over a system without precoding.

There is a growing need in the telecommunication industry for a scalable switch that can provide high-throughput communication between a large number of I/O ports: a terabit switch. Recent advances in the area of fiber amplifiers has spurred interest in `transparent' optical networks, wherein communication between users is achieved without multiple conversions between the optical and electrical domains. Moreover, polarization compensators have been developed for single-mode fibers to allow automatic and stable control of the polarization stage of output optical signals. This may enable a polarization-independent switching system that uses polarization-dependent `all-optical' switches. Polarization switching has been widely proposed in the context of free-space optical multistage interconnection networks for switching applications as well as for multiprocessor interconnections. In this paper we suggest a novel polarization-controlled free-space optical switch and present the implementation and characterization of a 4 X 4 photonic switch. The switching system is based on a unique optical element capable of acting with an arbitrary independent phase function upon illumination with horizontally or vertically polarized monochromatic light. This element, known as a birefringent computer generated hologram (BCGH) is composed of two birefringent substrates, etched with a surface relief pattern and joined face to face. BCGH optical interconnects provide arbitrary independent, efficient responses to the two orthogonal linear polarizations, thereby reducing the number of optical components in the free-space optical system.

The integration of photonic interconnects at the gate-to-gate level within the digital electronic environment has been demonstrably shown to significantly reduce the power consumption of the gate, and therefore, the circuit as well as the resulting system. This reduction can exceed two orders of magnitude when compared to current semiconductor implementations. Specifically, switching energies of 6 femto Joules (fJ) per bit have been achieved as compared to conventional implementations which range from several tens to several hundreds of fJ. Two fJ should be achievable with the correct balance of photonic and electronic technologies. HPOC modules with greater than 5,000,000 usable gates are projected to have sub-threshold leakage currents at < 10 nA and threshold currents of 1.25 (mu) A at a 1 GHz clock rate.

We present a storage scheme suitable for relational database operations such as projection, selection, join, and sorting. Relational database storage can be accomplished using optical memories such as two-photon 3D storage, volume holographic storage, spectral hole-burning memories, etc. Optoelectronic smart pixel arrays are used for data processing. The principles of the tomographic storage scheme and database operation algorithms are presented.

This paper describes an attempt to provide a practical application of a fourier optical signal processing as a preprocessor to a digital signal system. The input illumination is a coherent light source from a HeNe laser and computation of the Fourier Transform (FT) is carried out via an FT lens. An object with its symmetry is placed in the front focal plane of the FT lens and at the back focal plane of the lens, the Joint Fourier transform image results. Limitation and boundary condition of the optical system were studied and experimented. Problem encountered during transformation due to differing light source and intensity to various degree of coherency were also considered both at the optical processing and the digital system processor/compensator. Phase measurement and the digital processor/compensator uses the sum-of-product carry save add-subtract unit with reconfigurability so as to allow optimum cost-performance relationship. The sum-of-product consists of 24-bit floating point units capable of performing efficient multiplication, addition and subtraction. Many of such floating-point units are used in parallel in a pipelined architecture to perform Discrete Fourier Transform or other complex function for Digital Signal Processing computations. This also helps to understand the limitation and accuracy of the transformation and the practical behavior of the optical lens-camera system.

A general-purpose electronic computer is introduced to run a variety of parallel algorithms efficiently. Various levels of modular structures function independently and maximum flexibility results. Furthermore, the integrated modular components are all connected optically to each other to reduce the communication time dramatically. Computational power of the new computer is derived and the capability of handling intensive computation and complex information processing in real time is proved.

We present an integrated analysis of the effects of optical crosstalk and electronic noise on the BER performance of angular and spatially multiplexed volume holographic data storage systems. Our results show that increasing the SLM contrast ratio beyond 30 is of limited benefit in improving the system BER, and that electronic noise in the CCD is the dominant noise source when the signal is not shot noise limited (for ideal holographic materials). In addition, we show that CCD's with 100 or more low noise output channels are needed to achieve a 1 Gbit/sec readout data rate for reasonable preamp data bandwidths of 20 MHz or less.

We describe a page-formatted random-access holographic memory designed to store up to 160,000 holograms. The memory consists of 16 vertically spaced locations, each containing 10,000 holograms, which in turn are organized as 10 fractal-multiplexed rows of 1000 angularly-multiplexed holograms. A segmented mirror array is used to enable random access to any of the stored holograms within the access time of a non-mechanical angle scanner such as an acousto-optic deflector. Using a mechanical scanner with such a mirror array, we demonstrate storage of 10,000 holograms at a single location of the system, as well as simultaneous storage and recall of holograms at 6 locations, including the highest and lowest of the 16 locations.

Holographic data storage requires reference beam encoding for multiplexing and demultiplexing. Electrically switchable holographic composites (ESHC) based on a Polaroid photopolymer films are being investigated as a basis for several types of reference beam encoding devices. A laboratory demonstration reported here recorded four holograms in a photoresponsive storage medium. Both random phase encoding and angle multiplexing approaches were tested. ESHC encoding devices have much higher diffraction efficiency than spatial light modulators used in many other encoding schemes.

We describe a practical approach to 2D image storage in a coherent time-domain optical memory, which can be readily implemented with existing technologies. In this approach, images are stored spectroholographically in narrow (< 1 MHz) frequency channels of a time-domain storage material, with one image per channel. Advantages of this approach include fast single-frame recording time, variable playback speeds, random frame access, and the ability to perform in-memory image processing. Experimental results demonstrating high- resolution, high-fidelity, single-channel image storage with the use of a low-power, single- frequency ring laser are presented.

High-density and compact volume holographic memory is interfaced with electronic processors. Writing and reading are performed in parallel through high-degree wavelength and angular multiplexings. High-bandwidth of optical fibers are used to transfer data between memories and processors.

We discuss a volume optical storage, a 3D memory based on two-photon absorption process. This memory has an advantage over a 2D memory in that it combines fast access time together with high memory capacity--a capability that 2D memories cannot deliver. Two-photon 3D memory allows the data to be accessed in parallel, and thus speeds up data transfer rate significantly. In this paper, we address the issue of memory hierarchy which has been organized to provide a performance continuum of different memory technologies available at present. We justify the two-photon 3D memory as a viable technology to fill the performance gap currently exists between a primary and secondary storage systems. We also discuss the system capacity pertaining to two unique addressing schemes, i.e., orthogonal beam addressing and counterpropagating beam addressing. Finally, we report the progress in system study regarding the effect of memory material characteristic to the system design and development. Factors such as optimum wavelengths for read/write operations, material fluorescence, material fatigue, and the concentration of active molecules in the host material were considered.

The physical properties of the write and read forms of materials which may be used for permanent storage of information in 3D memory devices have been evaluated by means of picosecond and nanosecond spectroscopy. The writing and reading speed has also been determined.

Owing to unique optical properties the alkaline-earth sulphides doped with rare-earth ions find wide application as electroluminescent devices t1,2], .7.' radiation counters ES] as well as recording media of optical information carriers with the possibility of multiple information rewriting (4—7].

A volume photorefractive optical memory architecture can be integrated into 2D arrays of reconfigurable optical interconnect pixels for high performance digital processing applications such as high speed telecommunications and computing. This volume memory array will ultimately be implemented into a high performance optical computing module architecture. These modules use planar diffractive optical interconnect elements where fixed programs are stored. The insertion of volume photorefractive optical interconnect element technology will provide wavelength selective interconnect reconfigurability at system clock speeds (or faster). Target switching speeds of 1 ns may be achieved by electrical tuning of a multisection distributed Bragg reflection edge or a vertical cavity surface emitting laser diode. This photonic random optical memory access system is referred to as `PROMAC'.

A high resolution Mueller matrix imaging polarimetry test bed has been constructed and calibrated that has unique capabilities for characterizing optoelectronic devices, such as liquid crystal modulators, PLZT modulators, quantum well modulators, and surface emitting lasers. Similarly, the instrument can perform end-to-end measurements on optoelectronic systems including optical computers, interconnects, and correlators. It addresses, at the systems level, the need for incorporating polarimetric analysis and measurement techniques into the design, alignment, and testing of photonics technologies. The polarimeter maps the polarization altering characteristics of optical devices and optical systems, producing means of the retardance, the diattenuation, and the depolarization. The polarization mappings may be obtained across individual pixels or across large pixel arrays. The data sets provide a wealth of information not otherwise accessible for characterizing device uniformity, operating parameters, angular bandwidth, as well as identifying non-ideal polarization characteristics.

It is important to control the cross-talk that decreases the write/read window margin on the 1.3 GB - 130 mm MO disk, since the recording track is adjacent to the embossed pits of the header. We think this cross-talk is caused by two effects, one is the diffraction of the reflected light by the adjacent embossed pits and the other is `micro-birefringence' which exists in a local stress field around the embossed pits. We analyzed theoretically the relation between the micro-birefringence effects and macro-birefringence in a PC substrate, using a simple model of the micro-birefringence. The result shows that the cross-talk value depends on both vertical and in-plane macro-birefringences of the substrate. The combinations of the vertical and in- plane birefringences that make the maximum MO signal, don't make minimum cross-talk. Since the cross-talk is enhanced by the macro birefringence, only a narrow region of the vertical and in-plane birefringence are available to get the high quality of the MO signal-to- cross-talk ratio.

The results of the experimental investigation of the interaction process of focused laser radiation with metal-polymer films with uniform and gradient metal spreading are presented. A model construction of this process and focused laser radiation diffraction on pits are discussed.

The results of technical determination analysis are given; this analysis had been used while creating mass memory on optical cylinders. The distinctive features of accumulators on the cylinder carriers are considered.

The immersion system of information recording on optical cylinders is considered. The investigation results of system aberrations caused by the presence of cylindrical surfaces, immersion liquid by the change of such parameters as thickness, refractive index are presented. The recommendations on correction of such main system aberrations as spherical and astigmatism are proposed. The possibility of creation of high-aperture aspherical immersion objectives for information recording is shown. On experimental specimens of immersion-type information recording devices on optical cylinders is obtained the density of 8x11 It/mm.