We present a new class of non-linear correlation filters that produce arbitrary quadratic decision surfaces. These new filters first linearly combine the output from other linear or non-linear filters using complex-valued weights. These linear combinations are then passed through a square- law function and again linearly combined to produce these decision surfaces. The linear correlation filters are designed separately form the non-linear fusion parameters. The output from this new algorithm is thresholded to allow tradeoffs between probability of detection and the probability of false alarms to be made. This algorithm is numerically very efficient as it reduces the number of correlation operations required. It is optically and electronically implementable.

A design method for synthetic discriminant functions is described that optimizes filter performance by an appropriate weighting of the phase and of amplitude components of the training set images. The optimization criteria are the quality of discrimination of in-class and rejection of out-of-class images, filter efficiency and robustness in the presence of noise. The comparison is with the POF/fSDF. A practical demonstration is found in the application of a so optimized filter in a hybrid correlator.

COrrelation filters designed for automatic target recognition can be used for assisting trackers. We show that a tracker may use the filters for independent verification of the target as necessary. In particular, a form of the filters known as unconstrained correlation filters is shown to be optimal from a detection standpoint. An example is given to illustrate the concept of target reacquisition when the tracker's lock is broken.

Until now, most optical pattern recognition filters have been designed to process one image at a time. In contrast, many point-source target processing algorithms utilize successive frame integration to enhance the signal-to- clutter ratio. Our aim is to utilize the temporal correlation between successive frames in order to improve the tracking of extended targets appearing on very cluttered backgrounds. In our image model, the successive frames are assumed to consist of a moving object appearing on a moving background. From this model, the maximum-likelihood processor for tracking the object from one frame to the next one is derived. Given some simplifying assumptions, this processor is shown to consist in the linear combination of two sub-processors which are based on correlation operation. They could thus be implemented on a hybrid optoelectronical system that utilizes the rapidity of optical correlation.

A fully implemented vision based tracker must be able to identify an object in a variety of poses or distortions and estimate its position in a scene. After locking onto the object, continuous steady state tracking is required as the object gradually changes its position and orientation. The tracker must be able to recognize loss of track and take action to recover the object during break-lock and other transient conditions. While correlators can optimally recognize an object under theoretically ideal conditions, autonomous tracking of objects would require the development of a high level controller, or intelligent supervisor to deal will an uncontrolled visual environment. The supervisor would need to configure and control the analysis of the input environment, the detection procedure of the target, the trajectory estimation for maintaining lock on the target, and the camera orientation. In this paper we review the tracking problem. We then describe supervisor design based on configuring a suite of specific operations. Some of the operations include wide-area scan and prescreening operations using vector inner product composite filters; accurate detection and location with distortion-invariant composite filters to isolate a large data base of training views; filter banks distortion and which is distinguishable from characters associated with other objects. We discuss both hardware and algorithm considerations for the tracking problem. A general conclusion is that specific composite filter designs can be combined and configured to perform the tracking process.

Detection requires locating all objects in a scene independent of their class or aspect view and rejecting clutter. We consider new eigen filters to achieve this. They can be implemented on optical or digital correlators. Shift-invariance is required. These filters must also reject unseen clutter. Test results are presented for multi-class Synthetic Aperture Radar (SAR) data.

A intensity-invariant recognition technique was proposed for 3D object distortion-invariant OPR. A new preprocessing algorithm for training image set constructing and synthetic discriminant function (SDF) synthesizing technique were used to realize intensity-invariant SDF, and a input scenery image dynamic thresholding preprocessing technique was used to optimize the real-time optical correlation S/N of SDF and input scenery image. As result, the optical correlation S/N is independent of recognized object image intensity and distribution, in other words, the illuminating condition, reflectivity or thermal radiation condition of recognized object. An intensity, rotation, depression, scale and shift invariant OPR was realized simultaneously by using this technique with high S/N, sharp correlation peak, high intraclass recognition possibility and excellent interclass rejection ability.

This paper proposes a novel pattern recognition methodology based on morphological transforms and genetic algorithms. An entropy function is defined to demonstrate the match degree between two functions used in genetic algorithms. Based on morphological transforms and genetic algorithms, an optimal and adaptive set of structure elements as shape discrimination operators is developed by training patterns, moreover the string of variable structure elements is utilized to encode an image and construct the DNA of the image that maps arbitrary shapes into intrinsic and compact image features. Comparing the DNA string of the image with those of stored patterns, we can implement pattern recognition and classify an image. Then an optoelectronic pattern recognition architecture based on the algorithm is shown.

The projection-slice theorem, often used in tomographic applications for medical imaging, is utilized in conjunction with the SDF concept to implement a distortion-invariant filter. The marriage of these two well-known fields results in an effective tool for invariant pattern recognition. 1D filtering and phase-only techniques are used to implement the projection-slice synthetic discriminant function filter, which is then compared with bench-mark filters, such as the SDF filter and the matched filter in both complex and phase- only forms for a particular set of images.

Projective transformation (PT) is the most general transformation in nature. However, as it is nonlinear, people haven't made great progress in the research of projective invariant. On the other hand, many researchers are confused with PT and affine transformation (AT), they used projective deformed patterns testify affine invariants without any additional explanation. This paper has thoroughly studied PT and AT. When the fixed point or fixed axis of PT is on some special positions, PT degenerates into linear form - AT. In other cases, PT can be approximated as AT under certain condition. In this paper we derived a formula of relation between them and analyzed the errors arising from this approximation. It leads a new way to recognize projective deformed patterns. Three new independent affine moment invariants composed of second- or third-order central moments are adopted for 2D and 3D target recognition. The simulation experiment reached a satisfying result.

X-ray film and linescan images of pistachio nuts on conveyor trays for product inspection are considered. The final objective is the categorization of pistachios into good, blemished and infested nuts. A crucial step before classification is the separation of touching products and the extraction of features essential for classification. This paper addresses new detection and segmentation algorithms to isolate touching or overlapping items. These algorithms employ a new filter, a new watershed algorithm, and morphological processing to produce nutmeat-only images. Tests on a large database of x-ray film and real-time x-ray linescan images of around 2900 small, medium and large nuts showed excellent segmentation results. A new technique to detect and segment dark regions in nutmeat images is also presented and tested on approximately 300 x-ray film and approximately 300 real-time linescan x-ray images with 95-97 percent detection and correct segmentation. New algorithms are described that determine nutmeat fill ratio and locate splits in nutmeat. The techniques formulated in this paper are of general use in many different product inspection and computer vision problems.

A new method to detect an automobile registration plate using optical rank order hit-miss transform (HMT) is proposed. In this method, the optimal structuring elements are designed to recognize the corners of the registration plate and then they are applied to the optical rank order HMT. Optical rank order HMT gives better performance than the standard HMT in noisy or cluttered images. We present that whether the input images have random noise, clutter or not, the proposed method can easily detect automobile registration plates, and also when the detected plate have noise or clutter, they are presented after eliminating the noise or clutter from the detected plates by morphological opening and closing. Computer simulation shows the proposed method has good detection capability and we confirmed this by optical experiment.

Morphological transformation provides a powerful, nonlinear means of quantitatively analyzing data sets such as images. This technique has traditionally been applied to feature location or feature removal, as in noise removal. However, the technique holds some promise for fast object classification. By viewing the transformation as a neural network, proven training techniques may be applied to optimize the performance. The critical step in applying morphology is the design of the structuring element or shape of the filter. By casting the problem as that of object classification and by properly defining error functions, neural network training techniques may be used to optimize performance. In addition, this view of the procedure as a neural network allows the generalization of the technique to include sequences of filters, which correspond to multiple layer neural networks. an optical architecture is being considered to implement a sequence of morphological transformations, taking into account known principles and limitations of the optics and of neural networks, in order to perform a complex object classification task. Then the corresponding morphological filter parameter will be optimized using neural network training techniques.

The military has a requirement for small, low-power, low- cost pattern recognition systems that are capable of locating and identifying high value hostile targets. Miniature optical correlators can perform 2D pattern recognition at greater rates than digital platforms of equivalent size, power and/or weight. The patented miniature ruggedized optical correlator (MROC) can be built to meet the environmental, size, power, and weight requirements of military and rugged commercial applications, and at a cost that will permit wide deployment of the capability. The second version of the MROC correlator consists of a ferroelectric liquid crystal device in the input plane for high light efficiency and incorporates a reflective magneto optic spatial light modulator device in the filter pane for very high speed operation. The correlator has a volume of approximately 20 cubic inches. The MROC module, which includes all drive electronics and interfaces, is a 6U VME module that occupies 5 VME card slots. In this paper we will provide a brief review of the MROC construction and present sample results obtained from the MROC II breadboard. Initial tests demonstrated very high correlation levels, i.e. excellent discrimination, at pattern matching rates of 1920 per second on visible and simulated LADAR images of military vehicles and digital images of fingerprints.

For the last eight years, Lockheed Martin has been developing state-of-the-art optical processing systems. These systems are being developed to support a wide range of government and commercial applications including imagery exploitation, targeting, manufacturing inspection and medical imagery analysis. This paper intends to highlight the history, applications, systems and algorithms developed by Lockheed Martin.

This paper discusses an optical correlator interfaces to a FLIR and laser rangefinder to aid aircraft landing aboard an aircraft carrier. The purpose was to recognize aircraft and provide precision track in spite of the engine plume which is visible in IR images. Toward the end of the program, an opportunity arose to piggyback on tests of a Navy FLIR system, on board the USS Enterprise. The Navy's developmental FLIR and laser rangefinder were mounted on the carrier and provided excellent imagery with concurrent range data. The correlator performed a limited set of experiments at sea, tracking an aircraft from 8000 feet until almost touchdown. The challenges to the correlator we operation in a harsh environment and real time interfacing with other hardware. Real time range information controlled a series of filters in the correlator. The system fit into a standard panel rack and utilized remote alignment. The system operated during the chock of aircraft launch and landing, with no need to open up the optical box.

Optical processing systems often require compact high frame rate spatial light modulators (SLMs), usually with application specific modulation requirements in the complex plane. In this paper we discuss several advances at Boulder Nonlinear Systems toward this goal, including our liquid crystal (LC) on VLSI 128 X 128 analog SLM, and our multispectral hybrid incoherent to coherent converter.We also present the analysis of optical modulation possibilities when utilizing zero twist nematic and planar aligned chiral smectic LC on a reflective backplane. Finally we present the design of a multispectral optical correlator for machine vision applications such as food inspection, security, or manufacturing inspection.

Essex has developed the ImSyn processor, a sophisticated hybrid of optics and electronics. ImSyn calculates a discrete Fourier transform. The current production ImSyn is optimized for synthetic aperture radar processing, but has been used to process MRI, acoustic tomography, and synthetic aperture microscope data. The key feature of the production ImSyn is the ability to calculate images from non- rectilinearly gridded data. This data cannot be transformed with the FFT algorithm without interpolation or regridding. An alternative version of ImSyn is being developed for correlation applications. The correlator will be optimized for speed in performing rectilinear transforms.

A compact optical correlator based on binary liquid crystal spatial light modulators has been built. The correlator has a 21 X 28 cm² footprint and can process 256 X 256 pixel images at a maximum frame rate of 220 Hz. The system is insensitive to transportation and can be used both in VLC and JTC configuration. It can process live images from an external camera as well as images from computer memory. Variations of illumination conditions can change the graylevels and perceptible details of a target considerably. When this is the case, classical filters like phase only and optimal trade-off fail to detect the target. An improvement of the results can be achieved by using the optimum generalized filter. In most cases, the contour of the target remains more or less the same even if the graylevels are fluctuating. We show that by edge-enhancing and binarizing the input images used together with the optimum generalized filter we can achieve comparable results as compared to using grayscale images.

A compact, ultrahigh-speed, optical correlator using bipolar-amplitude-valued filtering is described. The system compactness and high speed is achieved by utilizing a newly developed nematic twist liquid crystal SLM by KOPIN as the input SLM and a ferroelectric liquid crystal SLM, optics design, and miniature packaging components. The utilization of a ferroelectric spatial light modulator (SLM) developed by Boulder Nonlinear System (BNS) in the Fourier transform domain of the optical correlator has enabled the bipolar- amplitude modulation for filter synthesis. The unique bipolar-amplitude modulation characteristics of the BNS SLM has made possible the design of real-valued spatial filters for pattern recognition. We have demonstrated the superiority of this real-valued filter for higher signal-to- noise-ratio correlation under cluttered environment. Moreover, the real-valued filter synthesis capability also allow us to extend the processing capability from mere optical correlation to optical wavelet transform, optical morphological processing and many other processing operations.

The recognition system rapid application prototyping tool (RSRAPT) was developed to evaluate various potential configurations of miniature ruggedized optical correlator (MROC) modules and to rapidly assess the feasibility of their use within systems such as missile seekers. RSRAPT is a simulation environment for rapidly prototyping, developing, and evaluating recognition systems that incorporate MROC technology. It is designed to interface to OLE compliant Windows applications using standard OLE interfaces. The system consists of nine key functional elements: sensor, detection, segmentation, pre-processor, filter selection, correlator, post-processor, identifier, and controller. The RSRAPT is a collection of object oriented server components, a client user interface and a recognitions system image and image sensor database. The server components are implemented to encapsulate processes that are typical to any optical-correlator based pattern recognition system. All the servers are implemented as Microsoft component object model objects. In addition to the system servers there are two key 'helper servers.' The first is the image server, which encapsulates all 'images'. This includes gray scale images and even complex images. The other supporting server is the filter generation server. This server trains the system on user data by calculating filters for user selected image types. The system hosts a library of standard image processing routines such as convolution, edge operators, clustering algorithms, median filtering, morphological operators such as erosion and dilation, connected components, region growing, and adaptive thresholding. In this paper we describe the simulator and show sample results from diverse applications.

Optical correlation systems have been lagging in performance due to the severely limited dynamic range capability which has been somewhat alleviated as newer devices are becoming available. In general, however, it is not clear what dynamic range is required for generally good performance both in the input device and in the filters. Experiments with a digital cross correlator which can accommodate up to 16 bits of input dynamic range was used to establish practical performance limitations for a majority of cross correlators. The experiments were conducted using a reduction of the dynamic range from actual 12 bit data incrementally to smaller levels such that performance measures could be compared. This paper quantifies these results.

A measure is proposed for the utility of a spatial light modulator, as relates to using the SLM in a realistic optical correlator. In this paper, the metric of utility is particularized to the birefringent liquid crystal SLM.

We investigate some methods of transforming an electronic video image prior to inserting it into an optical correlator. In order that the results might be implemented as a video transfer function, we do only point operations. We consider the input characteristics of the input SLM and laos the statistics of the input image. We model the correlator as noiseless, and we evaluate the transformations by tow metrics: correlation intensity and peak-to- correlation-energy.

Let a reference image and an input image that is magnified, rotated, and parallel-translated from the reference image be given. This paper discusses a method of computing the scale of magnification, the angle of rotation, and the quantity of parallel translation with a Hough transform of O'Gorman- Clowes version and a Fourier-Mellin transform. Differently from known methods using Fourier or Hough transforms, the discussed method can compute uniquely the angle of rotation. Moreover, the discussed method can process even low quality images since it does not require extracting feature points differently from known methods. Experiments were applied to actual images of bills.

A rotation, scale and translation invariant pattern recognition technique is proposed.It is based on Fourier- Mellin Descriptors (FMD). Each FMD is taken as an independent feature of the object, and a set of those features forms a signature. FMDs are naturally rotation invariant. Translation invariance is achieved through pre- processing. A proper normalization of the FMDs gives the scale invariance property. This approach offers the double advantage of providing invariant signatures of the objects, and a dramatic reduction of the amount of data to process. The compressed invariant feature signature is next presented to a multi-layered perceptron neural network. This final step provides some robustness to the classification of the signatures, enabling good recognition behavior under anamorphically scaled distortion. We also present an original feature extraction technique, adapted to optical calculation of the FMDs. A prototype optical set-up was built, and experimental results are presented.

For a flexible pattern recognition system that is robust to the input variations, a feature extraction approach is investigated. Two types of features are extracted: one is line orientations, and the other is the eigenvectors of the covariance matrix of the patterns that cannot be distinguished with the line orientation features alone. For the feature extraction, the Vander Lugt-type filters are used, which are recorded in a small spot of holographic recording medium by use of multiplexing techniques. A multilayer perceptron implemented in a computer is trained with a set of optically extracted features, so that it can recognize the input patterns that are not used in the training. Through preliminary experiments, where English character patterns composed of only straight line segments were tested, the feasibility of our approach is demonstrated.

We present a hybrid shape recognition system with an optical Hough transform preprocessor. A very compact design is achieved by a microlens array processor. Using incoherent light, the processor accepts direct optical input without any extra image converter being required. The microlens array processor is constructed of a crossed assembly of two low-cost plastic lenticular arrays and a Hough transform weight mask. It is integrated in a compact objective barrel, which is attached directly to a CCD-camera like a conventional camera lens. The system delivers one output signal for each of the 64 X 64 microlenses. The resolution of the microlenses and the weight mask results in an extremely high degree of parallelism. It corresponds to a connection of 4k inputs and outputs by 16M weights in parallel. The feature extraction tasks of lower computational complexity and the classification, which can be performed in real-time, are implemented as a neural network on a personal computer.

A distortion-invariant joint transform correlator, based on the fringe-adjusted joint transform correlator and the synthetic discriminant function concepts, is presented. Computer simulation results show that the proposed joint transform correlator peak and lower sidelobes compared with the classical joint transform correlator.

A comprehensive automatic target recognition (ATR) system using a wavelet transform based target detection preprocessor and a neural network classifier is described. A compact, high-speed optical wavelet processor with full gray scale filter capability, recently developed at JPL has been used for real-time target detection preprocessing. An innovative feature extraction algorithm using the Hermite Moments has been developed and used for neural net classification. The extracted Hermite Moment features, with their greatly reduced dimension and efficient representation, has enabled rapid neural training with test with very high classification and low false alarm rate. Experimental demonstration for face recognition and vehicle classification has been successfully carried out using this ATR system.

Image processing has become a topic of high relevance for automated product inspection in industrial manufacturing. A typical problem is the examination of structured surfaces of textiles to identify or classify defects. Product inspection under real-time conditions requires very powerful image processing systems which motivates the implementation of optical system concepts. In this paper we present a prototype of a coherent-optical processor which is used for the detection of defects in textile web images at video frame rate. After discussing the processor architecture and its underlying filter concept experimental results demonstrate the applicability of the system which is proposed to work as preprocessor in an opto-electronic image processing system.

Periodic defocusing of the crystalline lens leads to generalization of an image projected on the retina. In the process, the fine structure of the image is eliminated while informative fragments (IF), that is, brighter spots with abrupt contour change, are emphasized. The shape and arrangement of IFs may be utilized by a visual system in order to trigger saccades and form receptive areas. Results of theoretical and experimental optics studies imitating physiological processes are described. These results may be of interest to physiologists and can be used to develop anthropomorphic technical systems. A brief comparative analysis of other anthropomorphic systems proposed by A. P. Ginsburg, T. Podijo, D. I. Tomsit and others is given. A specific model of anthropomorphic robot representing a holographic correlator processing image by defocusing and by applying a set of spatial filters. These filters are constructed using a set of elementary images formed from two images, a straight stripe and a round spot, recognized by all living beings. Such a robot can be used for drawing letters and numerals from their written images; for classification of many similar images; for processing aerial photographs to determine boundaries between woods, various crops, etc. Input image defocusing is shown to be also useful for narrowing bandwidth in TV systems; for automatic loading of optical information after removal of noise.

A multispectral version of an image consisting of multiple wavelet components allows for more flexible feature extraction when compared to the use of one wavelet component. We showed how a multiple-input joint wavelet- transform correlator could be used for multispectral analysis of an input image. For m wavelet scales, m versions of the wavelet and m copies of the input image were generated using conventional optics that are used as inputs to a joint wavelet-transform correlator. The output consisted of 4m-1 correlation results, one of which is the desired output. The space-bandwidth product of the system is the same as for a conventional tow-input joint-transform correlator.

A new kind of the image representation - logarithmic hord transformation of the images is suggested. A structures of an optical-electronic processor, performing operation of LHTI and of an image processing computer system are developed. The analytical estimation of time expenditures in the system is evaluated.

A new joint wavelet transform correlation based technique is proposed for feature extraction such as detection of edges in an unknown input scene. We exploited a modified version of the Roberts and Sobel wavelet filters as the reference images for extracting the edges of an unknown input scene. The performance of the proposed technique using the aforementioned wavelet filters are evaluated and compared using numerical simulations.

Circular harmonic function (CHF) filters are used to determine the number of objects in a cluster in which multiple objects are touching and to provide estimates of the center of each object. The objects are agricultural products, pistachio nuts, and the sensor is X-ray. The nuts can have any orientation (hence rotation-invariance is necessary) . Each nut is basically elliptical, but shape, size and edge smoothness vary plus significant internal gray-scale variations are present; these variations are distortions. To detect such objects, we consider rotation-invariant RI-MINACE filters. To detect the separate correlation peaks, we use peaksorting with a window and minimum threshold. New filter design issues that arise in this application of rotation-invariant distortion-invariant touching object detection include: use of gray scale or binary images; the CHF order to use; selection of the reference object, the size of the filter, the MINACE control parameter c, and the size of the window and threshold in peaksorting. Keywords: circular harmonic filters, detection, distortion-invariance, rotation-invariance, rotationinvariant MINACE filters.

This study is an experimental analysis of a binary joint transform based fingerprint verification system. A constant false alarm rate performance analysis is used. Correct detection of matching prints is defined as a valid pass. Correct (non)detection of unmatched prints is a valid reject. Rejection of matching prints is a false alarm, and passing unmatched prints is a false pass. In this evaluation, the probability of false alarm is fixed by the detection threshold setting, and the resulting probability of false pass is analyzed. It is shown that for print rotations up to +/- 3 degrees a false alarm rate of 1 percent can be maintained with a probability of false pass of less than 5 percent. For a smaller acceptance window on the print rotation, better performance can be obtained. Experimental and simulation data is presented.

This paper presents results on direct optical matching of inked and real-time fingerprint images. Direct optical correlations and hybrid optical neural network correlation are used in the matching system for inked fingerprints. Preliminary results on optical matching of real-time fingerprints use optical correlation. The test samples used in the inked image experiments are the fingerprint taken from NIST database SD-9. These images, in both binary and gray level forms, are stored in a VanderLugt correlator. Tests of typical cross correlations and auto correlation sensitivity for both binary and 8 bit gray images are presented. When global correlations are tested on a second inked image results are found to be strongly influenced by plastic distortion of the finger. When the correlations are used to generate features that are localized to parts of each fingerprint and combined using a neural network classification network and separate class-by-class matching networks, 84.3 percent matching accuracy is obtained on a test set of 100,000 image pairs. Initial results with real- time images suggest that the difficulties resulting from finger deformation can be avoided by combining many different distorted images when the hologram is constructed in the correlator. Testing this process will require analysis of 10-20 second sequences of digital video.

We propose a new data protection method for the security system which controls the entrance of authorized persons. The basic idea is that the input image is encrypted by performing optical exclusive-OR (XOR) operations with the key bit stream that are generated by digital encryption algorithms. The gray level input image is converted to eight bit planes which are represented on a liquid crystal device (LCD). The key data represented on different LCD is reproduced to eight bit planes by lenslet arrays. The optical XOR operations between the key data and the bit planes are performed by polarization encoding method. The results of XOR operations which are detected by a CCD camera are converted to the encrypted gray level image and the image is used a an input to the joint transform correlator for comparison with reference images. We present some simulation results that verify this proposed method.

Optical random phase masks have been proposed for security applications and image encryption. A speckle noise problem caused by the random phase mask over the reconstructed image was analyzed in a previous work, and a system using speckle- free Fourier holograms and a phase-mostly spatial light modulator (SLM) was proposed. In this paper we propose a method for the optical implementation of image encryption using a speckle-free phase Fresnel hologram and a phase- mostly spatial light modulator. A speckle-free Fresnel information. Since this approach could be implemented using a reduced number of lenses or no lenses, it has a potential for commercial applications.

Designing a pattern classifier remains a difficult problem especially in the presence of noise and other degradations. Combination of multiple classifiers appears to be a good way of retaining the strengths of different classifiers while avoiding their weaknesses. Different combination schemes were proposed in the literature. As a special case of combining multiple classifiers, we consider combining correlators. Correlators are attractive for use in Automatic Target Recognition systems. Many correlation filter designs have been developed, each with its own features. Some filter designs maximize noise tolerance but do not provide sharp peaks. On the other hand, some correlation filters yield sharp correlation peaks but are overly sensitive to input noise. In this research effort, we explore the use of artificial neural network as a tool for combining correlators. Results of this implementation show improvements and indicate that combination of multiple correlators can potentially improve the classification performance.

Experimental investigating the possibility of optical processing through turbulent media were investigated and a model for turbulence is discussed. A technique for creating an optical device to display the turbulence model real-time is introduced and a technique for removing the turbulence is demonstrated.

This study demonstrates that the single-lens joint transform correlator is capable of position resolution accuracy on the order of a half micron. The single-lens joint transform correlator (JTC) is an extension of the chirp-encoded JTC in which the output lens has been eliminated. This is done by realizing that a Fresnel zone plate is formed in the joint power spectrum of the chirp-encoded JTC. This zone plate is illuminated with a plane wave and focuses to a point in the output plane. When chirp-encoding is used this way, it results in magnification of the output measurement. This is coupled with use of a higher order focus in the output plane to yield very high output resolution.

Efficient diffractive optical element (DOE) spatial image separator devices that segment an image and redirect corresponding regions to independent photodetectors are presented. These devices constitute a compact single-element alternative to prism arrays or reflective wedge arrays. DOE's with micro features are fabricated suing conventional microlithography techniques amenable to low-cost production. The system application described is a high-speed RF phase measurement processor for the NRL channelized direction finding receiver. This paper describes the design, fabrication, analysis, and empirical measurements of a 3 segment and 15 segment diffractive spatial image separator.

A phase-based processing technique is presented in this paper. In certain synthetic aperture applications, such as the ground probing radar, wide-band radar signals were utilized to detect the presence of targets. In these applications, amplitude information plays the major role at the signal processing stage because it seems less suffered from the complex environment. However, the phase of the echoes can also provide some useful information for the detection or location of targets. This paper attempts to use the phase information to provide a graphical perspective for the synthetic aperture application. Short-time Fourier transform was applied to analyze the phase-varying echo signals, the position-varying properties of target echoes can then be expanded on the position-frequency plane. The results show that the phase-based processing technique may be a promising approach to embedded targets detection and location.

In this paper, we propose two adaptive filters - the multiple median related (MMR) filter and the least-mean- square-error (LMSE) filter. The MMR filter can reduce edge drifting and preserve more fine detail than the median filter. The LMSE filter, which is based on the local signal- to-noise ratio (SNR), combines with MMR filter, yielding a filter that can remove non-impulsive noise. Our experimental results the proposed algorithm is robust in preserving edges,e reducing edge drifting and suppressing mixed noise in low SNR images. In addition, the algorithm can be implemented by parallel architecture, which enables real- time filter processing of noise-corrupted images.

Signal glow graph technique has been adopted to the task of calculating the transmission and reflection formula of light passing through a weakly absorbing three layer, or five phase system. It yields an exact closed form solution without neglecting higher order reflections between layer boundaries. Flow graphs have been solved by the analytical and topographical method leading to the same result.

Two variants of joint transform correlators with monochromatic spatially incoherent illumination are considered. The Fourier-holograms of the reference and recognized images are recorded simultaneously or apart in a time on the same spatial light modulator directly by monochromatic spatially incoherent light. To create the signal of mutual correlation of the images it is necessary to execute nonlinear transformation when the hologram is illuminated by coherent light. In the first scheme of the correlator this aim was achieved by using double pas of a restoring coherent wave through the hologram. In the second variant of the correlator the non-linearity of the characteristic of the spatial light modulator for hologram recording was used. Experimental schemes and results on processing teste images by both variants of joint transform correlators with monochromatic spatially incoherent illumination. The use of spatially incoherent light on the input of joint transform correlators permits to reduce the requirements to optical quality of elements, to reduce accuracy requirements on elements positioning and to expand a number of devices suitable to input images in correlators.

This article describes and compares some techniques to extract attributes from black and white images which contain musical objects. The inertia moment, the central moments and the wavelet transform methods are used to describe the images. Two supervised neural networks are applied to classify the images: backpropagation and fuzzy backpropagation. The results are compared.

An imaging optical system has been developed in order to undertake an experimental investigation of the importance of contact line and angle dynamics on the characteristics of dynamic liquid bridges. Liquid bridges, or captive drops, are fluid structures that occupy part of the gap between two or more solid supports. Experiments are performed in a neutral buoyancy tank. The optical system is a coherent high-magnification system which uses a high-pass spatial frequency filter to pass only the edges of the liquid bridge silhouette. System hardware design and realization, with special emphasis on alignment methods and aberrations present, will be shown. Sample experimental results will be shown as well. Potential methods for vibration mode recognition will be discussed.

A single step implementation of the joint Fourier transform correlation without any square law detector is proposed. Computer simulation results of the proposed correlator is included to verify its validity.

We have investigated a novel coherent optical correlator architecture for both recognizing and distinguishing between a plurality of patterns using only a single fixed binary phase-only matched filter, referred to here as a multiplexed filter. In order to utilize the filer space effectively we have introduced a novel spread spectrum technique for coding the input patterns, so called overlay encoding, suing a random high frequency binary phase-only mask. We have shown how standard optimization techniques can be used to design a multiplexed filter, which is capable of producing a set of codes in the output plane when the corresponding elements of a set of input patterns are presented in the input plane. The theoretical analysis has been verified experimentally using commercially available ferroelectric liquid crystal spatial light modulators.