We present several methods for optoelectronic processing of 2- and 3-D images based on digital holography. 2D objects are optically encrypted by using on-axis digital holography. The digital hologram is obtained with phase-shifting interferometry. The encryption of the hologram is performed by using random phase codes. As conventional holograms, digital holograms contain information about shape, location and orientation of 3-D objects. This allows us to apply 3-D pattern recognition techniques with high discrimination. We also measure 3-D orientation changes using the information contained in digital holograms. Experimental results are presented.

A new efficient two-stage algorithm to compute nonlinear features is described. Its implementation on a neural net with adaptive activation functions that raise the input data to an arbitrary power is described. Its use in face recognition with unknown input poses is presented.

In this paper we shall begin our discussion with the relationship between optics and humans, in which we see that light has indeed provided us with a very valuable source of information. A general optical communication concept is discussed, in which we see that a picture is indeed worth more than a thousand words. Based on Shannon's information theory, one can show that entropy and information can be simply traded. One of the most intriguing laws of thermodynamics must be the second law, in which we have found that there exists a profound relationship between the physical entropy and information. Without this relationship, information theory would be totally useless in physical science. By applying this relationship, Maxwell and diffraction-limited demons are discussed. And finally, samples of information optics are provided.

A consideration of the system advantages of a hybrid digital/optical correlator configuration together with some details of its construction were reported at the OPR X Conference. In this arrangement the input scene is digitally Fourier transformed at video-rate with a digital signal processor, the spectral data mixed with digitally stored templates and the result loaded onto a high frame-rate Spatial Light Modulator (SLM). An optical Fourier transform of this display then implements a correlation between the input scene and the reference template. Since the optical Fourier transform is practically instantaneous and the digital mixing and display on the SLM can accomplished at least two orders of magnitude faster than the digital Fourier transform, many reference templates may be searched during an input frame interval. Thus unconstrained pattern recognition problems may be tackled with this device.

Jet Propulsion Laboratory has developed a new 512 X 512 high-speed grayscale optical correlator (GOC) for real-time automatic target recognition (ATR) applications. As compared with a previous developed 128 X 128 grayscale optical correlator, the utilization of a pair of high-resolution input spatial light modulator (SLM) has increased the input field of view by 16 times. The use of a matching high- resolution filter SLM has increased the sharpness of the correlation peak. Key features of this GOC include: a grayscale input SLM (Kopin 640 X 480, 8-bit) to accommodate direct interface with the input imaging sensor. A real-valued (bipolar-amplitude) filter SLM (512 X 512, 4-bit) to enable use of a MACH (Maximum Average Correlation Height) composite correlation filter algorithm, compact and portable. This GOC architecture has greatly improve the system complexity by removing the need of preprocessing (binarization) the input, and the powerful MACH composite filter has greatly reduced the number of filter templates. In this paper, the criterion of selection of both input and filter SLM will be discussed. System analysis of building a compact correlator will also be provided. Experimental ATR verification of the 512 X 512 GOC will also be illustrated.

We have previously reported on the design and operation of a novel single-chip optical correlator prototype. Two ferroelectric liquid crystal SLMs and a high-speed APS camera were built into a single CMOS integrated circuit. Diffractive Fourier transform lenses were fabricated onto the surface of a window which was mounted on top of the chip. We are now working towards implementing the correlator as a business card-sized module mounted on a PCI card which can be plugged into the motherboard of industry standard PCs. We are also upgrading the SLMs to have analog optical modulation capability. The PCI card contains input and output image buffers, plus high-speed circuitry which digitizes the four analog output channels of the correlator's camera. This paper describes the system we are developing, some of the electronic and optical engineering issues involved, and the present status of our work.

Recent advances in our high-speed multi-level (analog) 512 X 512 liquid crystal spatial light modulator (SLM) will be presented. These advancements include smaller pixel pitch, greatly improved optical efficiency, and higher speed operation. The new VLSI SLM can utilize Ferroelectric Liquid Crystal to Nematic Liquid Crystal to achieve phase-only, amplitude-only, and phase-amplitude-coupled modulation. This device has applications in optical processing, optical storage, holographic display, and beam steering. Design criteria and experimental data will be presented.

A vanderLugt correlator is under construction using two Sanders binary Multiple Quantum Well Spatial Light Modulators. This correlation is part of the U.S. Army Aviation and Missile Command's three-year program to evaluate the latest generation hardware available for image processing. The Sanders modulators offer extremely high frame rates that make them attractive for image processing applications. The limiting factor in this correlator will actually be the output detector rather than the spatial light modulator. The correlator will be evaluated using infrared imagery to determine its performance in a target cueing/target screening applications. The results from these experiments as well as future plans will be presented.

Poor performance in key components has limited the development of optical target recognition systems. New components are now available, however, that exhibit size and speed characteristics compatible with image processing applications. The U.S. Army Aviation and Missile Command is well known for acquiring and testing a variety of spatial light modulators for image processing applications. We have recently acquired a pair of Sanders Multiple Quantum Well Spatial Light Modulators. These modulators will be evaluated in a manner similar to that used with previous modulators to determine contrast ratio, switching characteristics, optical efficiency, and residual phase modulation. The results of these measurements will be presented.

We report algorithms and laboratory practices that tell the full complex behavior of an SLM over its entire face, pixel by pixel, and put the information into a form that is useful to our filter optimization code. We add a quadrature component to the interferometry and image each pixel of the SLM. We analyze the fringes not at one value of drive and in an across-pixels dimension, but instead at each pixel in the drive dimension. We describe details of the method and given examples of spatially-variant filter SLM behavior. We provide examples of performance degradation when the filter's spatial variance has not been accommodated.

Current optimal filter projection methods either do not take explicit consideration of the limited dynamic range nature of the currently available SLMs or need complicated search and calculation process (for optimal gain and phase angle). To better utilize the limited dynamic range of the bipolar- amplitude SLM used in our recently developed grayscale optical correlator, we devise a simple and rather practical way to improve filter's dynamic range compression after using minimum Euclidean distance projection.

An efficient joint transform processor, called the phase- encoded fringe-adjusted joint transfer correlator (PFJTC) is proposed which employs phase encoding and nonlinear Fourier plane apodization for optimizing the correlation performance. For input scenes involving multiple identical targets, existing JTCs require multi-step on-line processing to eliminate the false alarms and other artifacts. The proposed PFJTC eliminates the possibility of false target detection by eradicating the false correlation peaks, and alleviates the effects of noise and other artifacts, in just one step thus ensuring higher processing speed. The phase encoding operation and the Fourier plane apodization does not have any detrimental effect on the processing speed. This technique ensures better utilization of space- bandwidth-product by using separate input and reference planes and by yielding only one peak per target instead of a pair of peaks produced by alternate JTCs. An all-optical implementation for the PFJTC technique is proposed and computer simulation results are presented.

Simulated annealing composite filter (SABCF) is a finite element spatial filter. It is trivial to assume that the number of training images to be stored in the SABCF is finite. In this paper, we have investigated the performance capacities for finite element SABCFs. We have shown that performance capacity increases as the number of pixel elements increases. We have also shown that the storage capacity of a SABCF is somewhat higher if the filter synthesis includes the anti-target training set. In other words, because of the suppression effect on the anti-target set, the SABCF would have a larger capacity as compared with the ones without using the anti-target training set.

The implementation of a compact Joint Transform Correlator designed for use in a production line validation system is presented. The application demands that the system is capable of capturing and examining printed product codes. The effects of applying non-linear processing techniques on the Input Plane and Joint Power Spectrum stages of the system are detailed, including morphological filtering. A subtraction technique is presented to further remove noise and unwanted cross-correlation terms from the Joint Power Spectrum. Installation of a CMOS camera into the system to replace the currently used CCD model is proposed offering a vastly expanded dynamic range. A technique for allowing the use of poor quality spectra is also presented which removes some of the sensitivity to noise and imperfections in the JPS.

Harmonic Wavelet Joint Transform (HWJT) has recently been proposed for automatic target recognition applications. The preliminary study shows improved noise tolerance for HWJT. In this paper, we investigate the application of JHWT in image de-noising. De-noising characteristics of this HWJTC have been studied and noise removal performances are demonstrated using HWJT correlator model for clutter image. To benchmark its performance, we compare its performance with the performances of well-established image de-noising techniques, such as Wavelet Packet. The image used for our simulation is a Synthetic Aperture Radar image from DARPA Moving and Stationary Target Recognition Program with angle of depression of 15 degree(s). Simulation results are presented to validate the performance of the proposed technique.

Correlation filters have been used very successfully in automatic target recognition (ATR) applications for ground- based tactical targets. Specifically, the maximum average correlation height filter has shown excellent performance in target detection and aimpoint selection for various types of armored vehicles in cluttered environments. This type of ATR is particularly well suited for implementation in an optical correlator due to the use of Fourier domain processing. We report preliminary test results of this type of SDF filter against ballistic missile defense targets and evaluate it in the context of real-time missile system performance.

In several recent articles it has been suggested that the shape of the correlation peak be used to distinguish between target and clutter. The peak shape is characterized in terms of some features, such as geometrical moments, which are then fed into a classifier that decides whether the peak was generated by target or clutter. The classification can be facilitated by an appropriate filter design. The maximum average correlation height (MACH) filter was designed to product similar correlation planes for target variations present in the training set. In this article we present generalizations of the MACH filter with the intention of decreasing the peak shape variation for targets in severe clutter. We show that by taking into account the non- overlapping character of the background noise and focusing the MACH correlation plane similarity requirement to the peak neighborhood, it is possible to simultaneously achieve a small variation in correlation peak shape and high peak- to-sidelobe ratios for cluttered images.

For the design of optically realizable correlation filters, two apparently different methods, the minimum Euclidean distance optimal filters and the multicriteria optimal trade-off filters have been proposed. In this paper, the equivalence of these two approaches is established by considering the operating regions that include the origin and figures of merit that include the detector noise.

This paper gives results for using distortion tolerant filters to improve performance of fingerprint correlation matching. Three types of distortion tolerant filters were tested: summation, weighted, and MINACE. A set of 55 fingers were used from NIST Special Database 24 to evaluate the filters. Our results show performance was improved from 49% correct, using one training fingerprint, to 100% correct, using multiple training fingerprints, and a distortion- tolerant MINACE filter, with no false alarms.

The projection slice filter is modified to include a noise cancellation algorithm for applications to real data with cluttered noise. The wavelet transform has been utilized in the color noise estimation process. The results of the application of these new techniques to the processing of real data with significant clutter components is reported. The techniques used here are part of an ongoing effort to improve the performance of correlation-based systems for fast recognition scenarios.

We have developed theory for computing filters with as large as Fisher ratio as possible. That theory analytically accommodates a number of real-world conditions, including noise or clutter in the input scene that is known to its power spectral density, additive noise in the detection process, and constrained filter values. The theory is adaptable to single-class pattern recognition. Using laboratory results we demonstrate Fisher-optimized filters that have improvements over some characteristics of our previous optimization of the Rayleigh quotient. To optimize a filter for the Fisher ratio is not free of side effects. We show examples of the penalty paid as one asks a filter to recognize more and more different objects.

When only certain filter pixels are allowed to transmit light, those `on' pixels constitute the region of support (ROS) for the filter. At those frequencies where noise power overwhelms signal power, or where the available values of the filter are not well aligned with the ideal spectral filter, the filter can actually work against the intent of the metric, and zero transmittance is preferable to what the SLM can do. In contrast to prior theories, which applied to only phase-only filters, we present algorithms for optimizing and studying ROS for an arbitrary SLM. We also present laboratory demonstrations of improvement in signal to noise ratio when a blocking mask is inserted into an optical correlator. A mask has detailed functional dependence on signal, noise power spectral density, the selected metric, and the realizable filter values. Rules are given for designing a mask that will be more broadly useful.

In this paper we demonstrate that dc-blocked phase-only filter correlation, based on image sharpening algorithm of SAR images (using an MSTAR data base) for both input and its matching template, is significantly improved. Our approach is tested with a very complicated image using a binary input and a binary matching template. Prior to binarization an image sharpening algorithm is used to enhance the input image and the matching template.

We propose a new approach for the 3D correlation. We compute directly the correlation integral without using the Fourier transform in the third dimension where the space bandwidth product of the sampled signal is small. We reformulate the 3D Fourier transform with a surface object model and the Lambert's Law. We discuss the means for recovering the object surface relief height function.

In this paper we introduce a new algorithm for an encoded filter for heterogeneous correlation by enhancing the cross- correlation between selected different objects. The new algorithm should allow the expansion of the use of correlation systems from recognition to classification. We tested the feasibility of this approach using a data base of stored information.

We present a simple theoretical model and compare it with experimental results for highly emissive sources having various coherence lengths in the presence of incoherent background illumination. Agreement between the experimental results and the theory is discussed.

A well known paradigm in scene reconstruction is the registration of a series of partially overlapping images into a larger image of the entire scene. A specific example of this paradigm is found in the reconstruction of aerial photography. Traditional correlation using frequency domain multiplication cannot efficiently register such imagery because the correlation requires a known template of the overlapping area, else the correct alignment response is corrupted by high sidelobe responses. We discuss a sophisticated supervised correlation process that makes use of special composite filter response characteristics to suppress these sidelobes. Thus, the detection and orientation estimation become independent of the amount, orientation and location of the overlapping regions. Our discussion will also include variations of this partial overlap paradigm that apply to a wide variety of applications in scene reconstruction.

In first part the Optical Pattern Formation system, i.e. the non-traditional Super-Scanning Locator, is shortly described. The goal of the second part--creation the Optical Pattern Recognition system working alike the Natural one, where receiving information, before reach the natural computer (brain), is pre-processed by the 5 organs of sense, mostly by analogous methods.

The problem of logo recognition is of great interest in the document domain, especially for document database. By recognizing the logo we obtain semantic information about the document which may be useful in deciding whether or not to analyze the textual components. In order to develop a logo recognition method that is efficient to compute and product intuitively reasonable results, we investigate the Line Segment Hausdorff Distance on logo recognition. Researchers apply Hausdorff Distance to measure the dissimilarity of two point sets. It has been extended to match two sets of line segments. The new approach has the advantage to incorporate structural and spatial information to compute the dissimilarity. The added information can conceptually provide more and better distinctive capability for recognition. The proposed technique has been applied on line segments of logos with encouraging results that support the concept experimentally. This might imply a new way for logo recognition.

Improvements on spatial light modulation made available cost-effective correlators suitable for industrial applications. One such application field is the real-time automatic inspection of textile web. Our approaches for defect enhancement make use of wavelet and Wiener amplitude- based filtering. Simulation and optical results are presented.

When the training class patterns {U_m} are closely related and they are inseparable (according to some targeted binary output vectors {V_m}) by a one- layered perceptron (OLP), then {U_m} must be linearly dependent, and the output-augmented {U_m} must be positively, linearly dependent. The learning of this OLP is then impossible no matter what learning rules we use, because the solution of the connection matrix just does not exist. However, we can always use a parallel-cascaded, two-layered perceptron (PCTLP) to realize this inseparable mapping. The design of this PCTLP is derived from the positive-linear independency condition we previously studied. It is a very intriguing mathematical derivation and the design of the PCTLP is much more efficient then that of the conventional series- cascaded, three-layered, neural networks. Also its robustness in recognizing any untrained, closely related patterns can be controlled and maximized. The physical origin, the theory, and the design of this novel, `universal' perceptron pattern recognition system will be discussed in detail in this paper.

A new image encoding and identification scheme is proposed for security verification by using a CGH (computer generated hologram), a random phase mask, and a correlation technique. The encrypted image, which is attached to the security product, is made by multiplying PCGH (phase CGH) with a random phase function. The random phase function plays a key role when the encrypted image is decrypted. The encrypted image can be optically recovered by a 2-f imaging system and automatically verified for a personal identification by a 4- f correlation system. Simulation results show the proposed method can be used for both the reconstruction of an original image and the recognition of an encrypted image.

Are presented the results of elaboration and investigation of the new class of the optical-electronic reconfigurable image processing computing systems (IPCS), based on the conception of the computing means with the architecture, controlled by the parameters of the input images. The systems are multiprocessor and refer to the class of functional-distributed. There is given a description of the basis model and 6 types of IPCS, based on usage of different optical processor, control computer. There are examined methods of computing processes organization in the system, presented results of the time expenditures. Comparative analysis of systems is given.

The report considers the evaluation of the automatic optical authentication technologies for the automated integrated system of physical protection, control and accounting of nuclear materials at RFNC-VNIITF, and for providing of the nuclear materials nonproliferation regime. The report presents the nuclear object authentication objectives and strategies, the methodology of the automatic optical authentication and results of the development of pattern recognition techniques carried out under the ISTC project #772 with the purpose of identification of unique features of surface structure of a controlled object and effects of its random treatment. The current decision of following functional control tasks is described in the report: confirmation of the item authenticity (proof of the absence of its substitution by an item of similar shape), control over unforeseen change of item state, control over unauthorized access to the item. The most important distinctive feature of all techniques is not comprehensive description of some properties of controlled item, but unique identification of item using minimum necessary set of parameters, properly comprising identification attribute of the item. The main emphasis in the technical approach is made on the development of rather simple technological methods for the first time intended for use in the systems of physical protection, control and accounting of nuclear materials. The developed authentication devices and system are described.

Nowadays different modifications of acoustooptic matrix-vector processor (AOMVP) are applied to the linear-algebra problems, spectrum analysis of radio signals, organization of reconfigurable interconnections, digital multiplication, neural networks' . One of the most promising AOMVP applications is parallel signal processing in synthetic aperture radar (SJ)2 Nevertheless, the advantages of optical information processing may be plagued because of the limited accuracy of computations in AOMVP, which are basically of analog nature. Therefore, theoretical and experimental investigation of computation accuracy is of considerable importance for SAR processor designers. Below the performance of AOMVP-based SA.R processor is analyzed in regard to the quality of fabrication, tuning, and adjustment of AOMTP elements, which is characterized with the totality of their noise and error root-mean-squares (rms's), as well as the value of crosstalk among the processing channels. The optical scheme of AOMVP experimental model intended for high speed processing of radar signals with using ofa multichannel acoustooptic modulator (MAOM) is presented

The scheme of the incoherent holographic correlator with two-coordinate acoustooptic deflector (AOD) for input of recognized images is presented. Because AOD forms an input image as a sequence of dynamic diffraction gratings, it allows to put a holographic filter directly behind the AOD. So, using this circumstance, the conventional scheme of incoherent holographic correlator was significantly reduced. The experimental setup of the correlator with holographic memory and monochromatic illumination is described. Experimental results on recognition of test objects with the semiconductor laser as illumination source are presented. The correlation signals are coincide with signals, received by means of computer simulation, within the limits of experimental errors. Results obtained shows that experimental setup of the incoherent holographic correlator provide real-time recognition of input images formed by two- coordinate acoustooptic deflector.

A new method to find five parameters of an ellipse from the optical Hough transform results is described. The method employs the Hough transform for detection of a straight line and the 1D analysis of the resultant parameter domain. By using this algorithm, we simulated about the ellipses with different positions, and we obtained the information of the ellipse with 94% accuracy in the worst case. To compare the simulation results with the experimental results, we performed optical experiments by using a HT CGH filter. Through the experiments, we showed that our results were very similar to those of the simulation results.

In this paper a new simple image encryption scheme and an optical decoding technique based on the principle of interference are proposed. An original image is encoded into a reference image and an encrypted image. The reference image is randomly generated, and the encrypted image is gotten by the encryption rule based on the principle of interference. These images have phase-only information and consist of two phase values `0' or `(pi) '. The interference image between two images becomes a binary image, which has a two-level intensity value. The performance of the proposed technique is evaluated using computer simulations and optical experiments.

Are presented the results of the modeling of the optical correlator. The efficiency of correlation functions calculation algorithms on the basis of amplitude-phase and phase-holographic filters for contour, whole objects' images as well as their Fourier spectrum modules have been investigated. On the bases of the obtained results can be proposed new methods of information processing.

Malignant melanoma is the deadliest form of cancer, fortunately, if it is detected early, even this type of cancer may be treated successfully. In this paper, we present a novel network approach for the automated separation of melanoma from benign categories of cancer, which exhibit melanoma-like characteristics. To reduce the computational complexities, while increasing the possibility of not being trapped in local minima of the back-propagation neural network, we applied Karhunen-Loeve transformation technique to the originally training patterns. We also utilized a cross entropy error function between the output and the target patterns. Using this approach, for reasonably balance of training/testing set, about 94% of correct classification of malignant and benign cancers could be obtained.

Electro-optical processors that employ cross-correlation or matched filter methods represent very efficient mechanisms that can be used for imaging sensor object identification. Such mechanisms are commonly referred to as optical correlators. Digital pre- and post-correlation processing steps are necessary to make the correlation operation robust. The pre-processing step conditions or filters the input image to create the most optimal image presentation for the correlation function. The post-processing step is typically a threshold operation on the cross-correlation of the matched filter with the optimal image presentation. The last action of the post-process is to determine whether to accept the threshold crossing as a match (identification) or dismiss the current matched filter and process any subsequent filers. A typical goal for the military application of the optical correlator is to maintain a high image frame rate of operation while maintaining an even higher matched filter selection rate. The goal is limited by the speed of the digital hardware and input/output bandwidth used in the pre- and post-processing. This paper addresses a novel methodology for using the correlator hardware with on- focal-plane array processing to maximize system throughput. The concept is described using the Image Algebra developed by Ritter, et al.