The incidence of breast cancer in women continues to increase1 . Studies have shown that early detection of breast cancer through periodic mammographic screening of asymptomatic women could reduce breast cancer mortality by 3050. However screening yields a high volume of mammograms requiring interpretation. In addition accurate characterization of detected masses is an important task of radiologists in order to reduce the number of unnecessary biopsies. Although some general rules have been suggested for the differentiation of malignant and benign masses34 considerable misclassification of masses still occurs. In fact on average only 20-30 of masses referred for surgical breast biopsy are actually malignant''5. As a potential aid to radiologists in mammographic screening programs we are developing a computer-vision system for the detection and characterization of masses in digital mammograms6''7. This system includes a detection subsystem and a characterization subsystem. Motivated by the systematic methods of viewing mammograms used by radiologists the detection system is designed to analyze the deviation from the architectural symmetry of normal right and left breasts and employs gray-level histogram analysis a bilateral-subtraction technique and run-length linking of multiple subtraction images to locate potential masses. False-positive detections are further reduced by various feature-extraction techniques. The characterization system employs various image analysis techniques such as the measurement of margin spiculation of masses in order to estimate the likelihood of malignancy.

It is of fundamental importance for biological scientists to assess cellular energetics. Under aerobic conditions, the tricarboxylic acid cycle (TCA cycle) is coupled with the mitochondrial electron cascade pathway to provide the cell with energy. The nicotinamide adenine dinucleotide-conjugated pair (NAD and NADH) is the coenzyme in numerous important biomedical reactions which include several important dehydrogenase reactions in the TCA cycle. Based on Le Chatelier's principle, NADH will accumulate when this energy production mechanism is impaired. The relative amounts of NAD and NADH in a cell are defined as the redox state of the cell (Williamson et.al. 1967) which provides a valuable index of cellular energetics. The sum of the amounts of NAD and NADH in a cell may be assumed to be constant during a finite time; therefore, a reliable means of measuring the NADH concentration would provide us with a useful indicator of tissue viability. Traditionally, the quantities of NADH and NAD may be measured by chemical assay methods. We can avoid these tediois analyses by exploiting the significant difference between the ultraviolet absorption spectra of this redox pair. However, because of the opacity of biological samples and the interference of other biochemicals that also absorb ultraviolet radiation, measurement of NADH and NAD+ concentrations in vivo by absorption spectroscopy is not feasible.

In Magnetic Resonance Imaging (MRI) selective enhanceitient of different features can be achieved by emphasizing the contrast of tissues with either long spin-spin relaxation time (T2) or short spin-lattice relaxation time (T1) . The contrast level is affected by the repetition tune (TR) and the echo time (TE) used in multiple spin-echo (MSE) iaethod. In this study anatomical structures of brain tissues are simulated by fractal geometry techniques and these images are then used to simulate MR images with different contrast levels. It is shown that the fractal grown brain slice gives a much closer resenthlance than a geoiaetric phantom and provides greater flexibility in simulating complicated structures. The fractal grown brain slice provides a computer generated phantom which is more suitable for simulating selective contrast enhancement in MRI. 1.

In this paper a new similarity measure based on the number of coincident bits in multi-channel images is presented. The similarity criterion incorporated in the image registration algorithm uses a coincident bit counting (CBC) method to obtain the number of matching bits between the frames of interest. The CBC method not only performs favorably compared with traditional techniques but also renders simpler implementation in conventional computing machines. An image registration algorithm which incorporates the CBC criterion is also proposed to determine the translational motion among sequences of images. Some experimental studies using low-contrast coronary images from a digital angiographic sequence have been performed. The results compare favorably with those obtained by using other nonparametric methods. Applications for this algorithm include digital angiography and mammography. 1.

We are developing an automated method for tracking the entire vascular tree in x-ray angiographic images. The vascular tree information which is obtained by our tracking method can be used for automated analyses of angiographic images such as detection and quantitation of vascular lesions identification of regions related to diseased vessels reconstruction of 3D representations from biplane images and analysis of blood flow. Our tracking method incorporates an efficient way of sampling the image data a connectivity test to assure that the tracking follows paths which are within vascular segments and a guided-search method in which information from nearby regions of the image is used to guide the tracking. Our current tracking method is more robust than a previous method and is capable of accurately tracking complex vascular trees which include tortuous vessels.

Visualization of liver image data in three dimensions (3-D) can reveal new information to aid physicians in volume calculation and surgical planning. Liver image data are first extracted from a series of 2-D cross-sectional abdominal images by using an automated segmentation technique. An interpolation technique is then applied to construct a 3-D liver image data set having the same spatial resolution in all three directions. This 3-D liver image data set is used as an input in surface rendering and volume rendering to generate two sets of 3-D liver displays. We have applied our methods to clinical liver images obtained from cases of a living-donor liver transplant program. 1.

The past few years have seen the growing use of digital techniques in the analysis of electron microscope image data. This trend is driven by the need to maximize the information extracted from the electron micrograph by submitting its digital representation to the broad spectrum of analytical techniques made available by the digital computer. We are developing an image processing system for the analysis of digital images obtained with a scanning transmission electron microscope (STEM) and a scanning electron microscope (SEM). This system, run on an IBM PS/2 model 70/A21, uses menu-based image processing and an interactive APL interface which permits the direct manipulation of image data.

Applying structural information from x-ray computed tomograghy (CT) to image reconstruction in positron emission tomography (PET) for compensation of the effect of photon attenuation can provide improved quality of the reconstructed PET images. CT and PET scans can be registered precisely using a surface-matching technique. Anatomical maps are then obtained by reslicing the CT images along the PET image planes. Attenuation coefficient maps for the corresponding PET images can be derived from the anatomical maps by considering the difference in photon energies observed in CT and in PET. These attenuation coefficient maps are reprojected to produce a set of attenuation correction factors for each projection profile. PET images reconstructed using these attenuation correction factors exhibit significant reduction in noise. 1.

The field of holography conipletely depends on the coherence of laser light. The most meaningful attribute of coherence for holographers is the coherent length. Commonly coherent length is regarded as the maxiinurct optical path difference between two bearas such as those in a Michelson interferorneter that results in a discernable interference pattern. Herein we discuss two direct holographic methods of measurements each yielding visually observable results. We begin by examining several theoretical models for laser spectral linewidths and their subsequent derivation of coherent lengths. 2. Spectroscopy and Line Shape Let us first discuss some basic tenants of laser spectroscopy since it leads us directly to the understanding of coherence. When the effects of the apparatus is progressively reduced and infinite resolving power is being used it is found that each spectral line consists of an intrinsic distribution of frequencies about the. line center. The profile of every spectral line has a finite width and characteristic shape which are determined by the conditions existing in the source and it is this profile which we now wish to consider. The line profile is important in determining many of the characteristics of gas lasers especially its coherence length. I. The Natural or Radiative Line Shape - the width of spectral lines errtitted by isolated stationary atoias. They are Lorentzian and honiogeneous. (i) Classical Model: Introducing the radiation reaction force it is easy to show

The philosophy of the holography department The School of the Art Institute of Chicago (SAIC) is discussed along with a fine arts curriculum designed to take into account the learning styles of art school students. A description of the facilities follow ending with a survey of artwork by the best students. Introduct ion Holography is almost unheard of in the art world when it is mentioned it is apt to be used as a metaphor for the ultimate in tastelessness. Occasionally a ''recognized'' artist will dabble in holography using the facilities of a commercial lab translating his concept into a hologram1. The results usually raise objections from the clique of handson holographic artists who feel that they are213 on the other hand are being discriminated against by the art world because they practice in a technical medium. As a way of improving the situation art school students must be better educated in the fine arts aspect of holography laboratory techniques and how to set up their own studios. For it will be only through the efforts of the " hands-on" workers will the medium fully realize its creative potential. This then is the major goal of this teacher: to help the students realize their full potential creating artful holograms with inexpensive equipment. Not only is it necessary that the holographic artist learn to work within the constraints of equipment at hand

X-ray lithography is a proximity printing process. After a resist-coated wafer is positioned 10-40 m behind the x-ray mask, the mask and wafer are aligned and secured for the 'x-ray exposure. Securing the mask may create distortions in the membrane. The lx nature of proximity printing demands that these distortions be tightly controlled. At O.25,tm critical dimension, the distortion budget allocated to the mask is 25 nm. This corresponds to 0.7 ppm over a field of 1" . It is thus necessary to develop a metrology tool capable of accurately and rapidly measuring distortions induced in the mask. Interferometry, because of its non-contact nature and high sensitivity, is ideally suited to the task. The construction of an x-ray mask begins with the deposition of a 1-2 m film onto a silicon wafer substrate.' After carrier deposition, the mask wafer is bonded to a glass or silicon mounting ring. The mounting ring gives the mask rigidity and provides a mechanism for mounting the mask during x-ray exposure. Mask distortions may occur in the bonding process due to non-fiat wafers or bonding ring irregularities. The central region of mask blank wafer is then back-etched with KOH to create an x-ray transparent membrane. The redistribution of stress in the carrier during the membrane construction is a second source of distortion in the mask-making process. The mask blank is subsequently patterned and metallized. Localized stresses occurring at the absorber-membrane boundaries can lead to additional mask distortions. Several interferometric techniques can be employed to characterize the sources of the above-mentioned x-ray mask distortions. Out-of-plane distortions (OPD) of x-ray masks have been measured with the Michelson interferometer. In-plane distortions (IPD) in the mask have been characterized using Moire interferometry. In Moire, diffraction gratings printed on the mask pre-exposure are examined post-exposure in a virtual grating. Both OPD and IPD have been observed in-situ using shear interferometry.

Various methods of imaging through inhomogeneous media and small apertures are described and experimental results are given. 1.

Endoscopic holography or endoholography combines the features of endoscopy and holography. The purpose of endoholographic imaging is to provide the physician with a unique means of extending diagnosis by providing a life-like record of tissue. Endoholographic recording will provide means for microscopic examination of tissue and in some cases may obviate the need to excise specimens for biopsy. In this method holograms which have the unique properties of three-dimensionality large focal depth and high resolution are made with a newly designed endoscope. The endoscope uses a single-mode optical fiber for illumination and single-beam reflection holograms are recorded in close contact with the tissue at the distal end of the endoscope. The holograms are viewed under a microscope. By using the proper combinations of dyes for staining specific tissue types with various wavelengths of laser illumination increased contrast on the cellular level can be obtained. Using dyes such as rose bengal in combination with the 514. 5 nm line of an argon ion laser and trypan blue or methylene blue with the 647. 1 nm line of a krypton ion laser holograms of the stained colon of a dog showed the architecture of the colon''s columnar epithelial cells. It is hoped through chronological study using this method in-vivo an increased understanding of the etiology and pathology of diseases such as Crohn''s diseases colitis proctitis and several different forms of cancer will help

We present a general method to generate high-fidelity phase conjugation from holograms by the use of a phase-conjugating (PC) mirror. Diffraction-limited phase conjugated images were obtained in our experiments. By applying this technique image transmission through multimode fibers can be accomplished. Such a holographic phase conjugation technique may have more applications for image correction through other aberrating media high quality image projection parallel data communication through short optical multimode fibers holographic couplers high quality laser beam delivery HOEs etc. 2.

Holographic interferometry is a useful technique for studying the vibrational modes of both separate vibrating elements and complete instruments. Spatial resolution is excellent and studies can be made at small amplitudes. Time-average holographic interferograms of handbells Chinese two-tone bells Caribbean steel drums snare drum shells cymbals and guitars are shown. 1 .

The fracture process on cement based materials was studied using various holographic interferometry methods. Several aspects of the fracture process were studied on plane mortar (cement water and sand) and polypropylene fiber reinforced cement under different loading configurations. Sandwich holography was used to analyze mortar specimens loaded in tension. The real time method was used to determine the load increments required for accurate and easy evaluation of displacements. A single sensitivity vector setup was used for the evaluation of crack opening displacements and a multiple sensitivity vector set-up was used for the evaluation of full displacement fields. In both cases the holographic images were acquired by a digital image analysis system. First the images were enhanced for better recognition of the fringe patterns by elimination of high frequency noise and contrast improvement. Then the image analysis system was used for automatic fringe count. Sandwich holographic interferometry was also used to map the crack surfaces of the mortar specimens tested. The use of image analysis was essential for the evaluation of the interferograms. The image analysis system allowed for the direct application of a moire pattern method resulting in topographic contour maps. Crack pattern characterization on polypropylene fiber reinforced specimens was made using the ''piggy back'' method to record double exposure reflection holograms. Image processing was used for image enhancement and automatic computation of crack spacings. 1.

A real-time holographic interferometry system was developed for the inspection and engineering analysis of disposable reagent containers used in a clinical chemistry analyzer. This system was based on a conunercial holographic camera and permitted the manual or automated detection of bonding flaws which could result in field failure. L

The use of a TV camera as a recording medium and the observation of whole field displacements in real time makes holographic TV a very interesting and powerful tool in a variety of areas from NDE to research and development. The paper presents new developments in the field that add to the versatility of the technique by introducing portability and methods to obtain accurate quantitative results. Examples of applications are given to the measurement of strains both at room and at high temperatures and strain measurements at the microscopic level. 1.

In this paper we discuss a fiber optic phase measuring Electronic Speckle Pattern Interferometer. The phase stepping is obtained by stretching the reference arm fiber. We also outline briefly the theory of determining the phase by differentiating the interferogram. Preliminary results obtained with four bucket and differentiation algorithms are presented. 2.

A study using holographic interferometry (HI) to examine human body surface motion has been done. Skin surface motion resulting from the pumping action of the heart is detectable with holographic methods. We have examined the skin motion in the neck area in detail. The interferograms obtained using a double pulsed ruby laser system provide an image of the human subjects with a fringe pattern superimposed upon that image. The fringe patterns correspond to the motion of the skin during the time between the two laser pulses. These fringe patterns were analyzed and correlated with several known cardiac phenomena. The patterns show a high degree of intra- and inter-subject consistency for healthy male subjects. To determine direction (sign) of skin displacement from standard interferograms a method of introducing reference fringes was incorporated into the system. To confirm that the fringe patterns yield accurate displacement data a point sensor was utilized to directly measure skin movement. Holographic interferometry''s future value as a clinical tool warrants further detailed investigation. 1.

This paper gives a brief overview of neural networks noting the advantages of the optical implementation. The Hopfield model mainly used as an associative memory is introduced. Planar holograms have been used to defme the interconnections between neurons in optical systems but volume holograms with photorefractive crystals offer much greater potential and density for these interconnections1. Because of the lack of suitable optical material light modulators have been used to simulate neurons in the neural plane. A simple one-layer network using photorefractive material will be presented along with some preliminary results.

We investigated a new concept for phase conjugate one-pass distortion correction. The distorted input is imaged into a BaTiO3 crystal. A phase conjugate is produced through a self-pumped phase conjugation technique. The nonlinear material acts as a time averaging filter. Furthermore for a time-varying distorter the self pumped phase conjugator removes distortion in real time. This technique requires only one pass through the distorting medium. 1.

Lithium niobate crystal has been used for storing information that can be accessed through an associative recall. This fact demonstrates the application of the crystal to the optical memory. specifically In an optical ring resonator. One important factor that determines the recall ability is the diffraction efficiency. Particularly this quantity Is also a function of the polarization of the readout beam. The diffraction efficiency of the crystal Is significant If the sufficient build-up of power in the real-time programmable ring resonator is required. To this end this paper is concentrated on the study of the dependency of diffraction efficiency on the polarization of recall beam. Experimental results are also given to indicate the significance of orientation of c-axis of the crystal in the experimental configuration. 2.

As more supporting theory and applications for neural nets are being developed the hardware implementation of neural nets becomes more important. However most of the existing hardware implementations of neural nets suffer from various weaknesses. Such problems range from inadequate support for parallelism to high communication overhead. In this paper we propose the design of a dataflow-based multiprocessor system for efficiently implementmg asynchronous highly parallel neural nets. L

An architecture for VLSI/ULSI implementation of neural network models is presented. The architecture is of the linear and systolic type and can easily be expanded to a larger network size with more processing elements per layer and/or more layers per network model. The weights in this architecture are fully adjustable. The weights must be pre-computed with the help of a host computer and downloaded to the system as a preinitialized process. This architecture is basically suitable for feed forward neural network architecture. 1.

The paper proposes a neural network model for two layer channel routing. We hope this research will lead to a better understanding of the capability and limitations of neural networks as a general design methodology and in particular when it is applied to the routing problem in the design of VLSI chips. In our model the neurons form a two dimensional array where the value of element N represents the " chances" of net i is positioned at track j. The strength of connections between neurons are defined such that to prevent horizontal conflict between nets and also optimize a number of important routing metrics such as: minimum routing area and minimum wire length. We ran several examples for most of the small size examples ( less than 1 5 nets ) we were able to obtain good solutions. However for larger size examples apart from the known problems of long simulation times the router was not able to find a solution in many cases. 1.

Influence of two new nonlinearity functions on Perceptron performance is studied. The two new functions under consideration are Gaussian and sinusoid functions. The new functions create multithreshold Perceptions capable of handling both binary and analog inputs. A computer program has been developed to simulate behavior of a network utilizing either of the two modified Perceptrons. Both XOR and Parity Check problems were solved using a single-layer network utilizing these modified Perceptions. Based on the results obtained from the simulation the modified Perceptions are capable of solving problems (such as XOR) that can not be solved using a single-layer of the classical Perceptron. Also networks utilizing these modified Perceptions require fewer number of iterations to converge to a solution than that of a multi-layer network of classical Perceptions using back propagation. In addition the results show that Sinusoidal Perceptronperforms better than Gaussian Perception. 1.

A new multi-threshold Perceptron capable of handling both binary and analog input is presented and discussed. The modified Perceptron replaces the sigmoid function with sinusoidal function. A computer program has been developed to simulate behavior of a network utilizing the modified Perceptron. Both XOR and Parity Check problems were solved using a single-layer network utilizing this modified Perceptron. Based on the results obtained from the simulation the modified Perceptron is capable of solving problems (such as XOR) that can not be solved using a single-layer of the classical Perceptron. Also a network utilizing this modified Perceptron requires fewer number of iterations to converge to a solution than that of a multi-layer Perceptron network using back propagation. 1.

A class of learning techniques for neural networks can be considered as optimization problems. The connection strengths are modified such that the difference between the network response and a desired response Is minimized. In this paper the learning techniques based on the gradient momentum Newton and quasi-Newton methods are considered. A learning algorithm is also developed based on the conjugate gradient technique. These learning techniques are applied to the Exclusive-OR problem for comparison of their performance. For this problem the algorithm based on the conjugate gradient technique converges faster than the other algorithms. 2.

The problem of collision-free path generation for mobile robots is considered This task is formulated into an optimization problem for which a recurrent neural network similar to Hopfield''s is useiL The outputs of the units in the network correspond to the position of some intermediate points on the path which would be connected by straight lines. The steady-state values of the outputs of the units in the network correspond to the coordinates of the position of the intermediate points on the shortest collision-free path to the targets To derive the dynamics of the network first an eneri function corresponding to the distance from the target the total length of the path and the degree of closeness to the obstacles is derived1 The dynamics of the network is then derived according to a gradient descent algorithms 1 .

The problem of distributed detection of a deterministic signal with unknown strength using a neural network approach is considered. Since the signal level Is not known there does not exist a uniformly powerful detection scheme for the problem. The neural network architecture used Is a two layer feedforward perceptron model with N Inputs M nodes In the hidden layer and two output nodes that correspond to the signal and noise hypotheses. The back propagation algorithm Is used to train the system with a known noise distribution. For fixed false alarm probabilities the performance of the trained network with regard to a set of test signals is estimated and compared to the performance of some known detection procedures such as Fisher''s method Tlppett''s method and the optimum test. Simulation results are presented for a network with two input nodes and three or four nodes In the hidden layer. When Laplace noise Is assumed the network performs better than the Tlppett method. For the detection of weak and moderate signals in Gaussian noise the network performs reasonably close to the optimum detection scheme. The decision regions for the neural network detection scheme is comparable to those of the Fisher''s and Tlppett''s methods. However unlike the Tippett''s method which has a fixed decision region regardless of the noise distribution the neural network''s decision region changes with the noise distribution. 1. SIGNAL DETECTION

Although neural networks are very effective pattern classifiers a major limitation is that they are not suitable for classifying patterns with Inherent time-variations. This paper describes an approach to incorporate a temporal structure in a neural network system which wifi accomodate the time variations in local feature sets encountered in problems such as partial shape classification. 1.

A neural network approach to power system's static security analysis Is presented. This security analysis includes both security assessment and enhancement. For security assessment, a three-layer feedlorward network has been developed for a small 5-bus system. This network represents the nonlinear nature of the power system accurately enough to identify severe security violations. The momentum method was used to train the network. For security control(enhancement) of power systems, a recurrent network approach Is proposed to replace the existing expert system technique. Given an insecure operating state, the proposed approach will produce the operating points which result in a secure operation of the system. The preliminary results Indicate that neural network approach (given neural network hardware availability) can be used for on-line security analysis in a power system's control center enviroment.

This paper discusses a dynamical network for mapping of interciass members without performing a learning process. This allows a member of class A to be mapped to a member of class B. Given sample members of each class a backpropagation network is trained to form the corresponding class boundaries. Upon completion of the training process the weights obtained are used in a recurrent network which performs the interclass member mapping without any further training. This mapping is achieved as the recurrent network evolves In time. The Initial state of the network is mapped to its equilibrium state. The interclass member mapping network (IMMN) has many applications in selfcorrecting systems. In this paper the IMMN is developed to represent two classes namely class B (for instance a class for representing members with desirable and correct features) and class A (members with incorrect features). An example is given in which two categories are used namely poorly and well-designed manufacturing parts. Given a poorly-designed part the network wifi suggest corrections resulting in a well-designed part. This example has nonlinear decision regions and shows the generalization capability of the network.

The construction of the mechanical and electronic components for the sextupole-corrected subangstrom resolution scanning transmission electron microscope (STEM) has been completed. They are now in the process of being tested. The required software for the control measurement diagnosis and operation of this instrument has been written. We have successfully obtained a beam from the field emission source and have achieved stable emission current.

Using a scanning ion microprobe we analyzed the distribution of labelled thymidine along human chromosomes. Two labels have been used: bromodeoxyuridine (BrdU which contains one bromine atom per molecule) and 14C-thymidine (which contains either one or ten 14C atoms per molecule). Both types of labelled nucleosides can be detected with our insirument. Best results are obtained when using the uniformly labelled thymidine (U-14C-thymidine) and adding up in a KONTRON IBAS image processing system the sequential analytical maps acquired from the sample at mass 28 (14C14N ions). The distribution of thymidine is heterogeneous along the chromosomes and a banding pattern can be observed on the pictures (SIMS-bands). The spatial resolution obtained with our scanning ion microprobe (the University of Chicago Scanning Ion Microprobe: UC-SIM) surpasses the one of autoradiography which is the common direct method of localization of labelled nucleosides. 1.

We have combined x-ray crystallographic coordinates of sickle hemoglobin (HbS) molecules with electron microscopic 3-D reconstruction data of HbS fibers to calculate residues involved in intermolecular contacts within the clinically relevant HbS fiber. The model accounts for the action of 80 of the fifty five point mutations of HbS whose effect on fiber formation has been studied. For instance fiber formation is affected by thirty-three mutants. Of these mutants thirty-one occur at the calculated intermolecular contact sites. Fiber formation is not affected by 28 known HbS mutants 16 of these mutations occur at residues not involved in intermolecular contacts. The 12 mutations that occur at contact sites are postulated to not significantly increase fiber formation energy. The number of intermolecular contacts decreases with radius until at the periphery of the particle there are no contacts. We suggest that this observation accounts for the limited radial growth of the particle and that a similar mechanism may be a factor in limiting the size of other helical particles.

Noncoherently detected transillurnination light waves only provide information about the absorption characteristics of a tissue subject. Coherent detection is capable of providing information about the dispersion characteristics. However a relationship exists between absorption and dispersion at optical frequencies. Both types of detection are seen to be mathematically equivalent. The difference lies in noise rejection capability and ease or difficulty of system realization. Models of absorption and dispersion properties are. also presented. 1.

As the networking of computers becomes more ubiquitous the need to connect imaging instruments to such networks becomes a matter of greater importance. The proliferation of personal workstations with graphics capabilities in particular is putting image processing power into the hands of an increasingly large number of researchers. By connecting image acquisition systems to a local or wide area network image data which once could only be analyzed in the lab in which the imaging instrumentation was situated may now be accessed in any lab equipped with appropriate image processing software and hardware for example image data may be remotely accessed from many modem state-of-the-art telescopic observatories.

We have developed a stand-alone image processing system on an IBM 6000-530 workstation to analyze digital images from our scanning transmission electron microscopes (STEM). The system is UNIX based and utilizes the X-Windows system for display of images.

As pressure on our limited land base continues to increase managers of public lands must have more accurate information within a shorter time to make logical defensible decisions which are acceptable to the public. Remote sensing technology provides many tools required to gather much of the information used by decision makers. Some of the most important remote sensing tools are based on laser and electro-optical technology. This paper provides an overview of some applications of laser and electro-optical devices by managers of natural resources. It is important for workers in other fields to be aware of the problems and needs of resource managers as it is important for resource managers to be knowledgeable about developments in technical areas. Sharing information will promote opportunities to develop new tools and improve the effectiveness and efficiency of management. Personal knowledge and literature searches provide examples. While the variety of uses in somewhat limited their importance is increasing as managers and analysts become more accustomed to using products of this technology. Lasers and electro-optical instruments will continue to be a very important part of our data collection process. 2. 0

The design and operational characteristics of the major types of photoelectronic detectors presently being made at our facility are discussed. These include vacuum photodiodes microchannel plate photomultiplier tubes proximity focused image tubes x-ray image converter tubes image tube intensified self-scanned array TV cameras and streak tubes. Other types of specialty detectors that have been made are also discussed as examples of the breadth of technology required in the photoelectronic detector field. 1_.

Veiling glare is a phenomenon present in most imaging devices and systems. It is a source of concern in image intensifiers where it causes a loss in image contrast especially at low spatial frequencies. Veiling glare measurements are described for two different F4lll 18 mitt proximity focused channel intensifier tubes (image intensifiers) one with a fiber optic input window (FO/FO) and the other with a quartz input window (Q/FO) . Near field veiling glare a raeasure of the stray light close to an edge and far field veiling glare a measure of more widely distributed stray light were measured as fractions of the illumination level. Near field veiling glare of 0. 3 0. 06 and far field veiling glare of 0. 02 0. 004 was measured for the FO/FO intensifier and near field veiling glare of 0. 3 0. 06 and far field veiling glare of 0. 03 0. 004 was measured for the Q/FO intensifier. Characteristic lengths a measure of the extent of the veiling glare from the edge are 0. 7 0. 07 mm (near field) and 2. 0 0. 2 mm (far field) for the FO/FO intensifier and 1. 2 0. 1 mm (near field) and 7. 0 0. 7 mm (far field) for the Q/FO intensifier. Studies also show that characteristic lengths increase with increasing wavelength. Further a quasi-elastic electron scatter model suggested that electron scatter within

A method of preparation and deposition of polymethyl methacrylate (PMMA) thin film optical waveguides is presented. This method is devised to produce low loss optical waveguides. Since the processing conditions in the film deposition are crucial to minimizing the optical loss and since they affect the optical property of the films many different factors involved in this process will be identified and their effects on the resultant films will be explained. Finally the modal structures of some typical films will be analyzed and the resultant attenuation constant will be given. 1 .

We have performed a study of the coupling of single mode optical fibers to integrated optical polymer waveguides. Present methods of fiber-waveguide---fiber coupling are discussed. The technique of fiber end polishing and its importance is discussed. The optimum geometry of the groove is predicted. Different possible losses and their remedies are described. The precise geometry of the V-groove removes all but two degrees of freedom in aligning the fiber to the waveguide. A further improvement in the etching technique can reduce this further down to only one linear degree of freedom. I .

A technique was developed to polish the tips of standard communication optical fibers at angles ranging from 35 to 55. This study was initiated to determine the feasibility of making right-angle fiber optic (elbow) connectors. The process used to polish the fiber ends at various angles and the experiments performed to characterize the light output will be discussed. A simple ray-optic computer model was used to analyze the measured light output profiles from the polished ends. 1.

This paper discusses an application of artificial neural networks in smart structures with fiber optic sensors. Emphasis is on using a novel neural network approach to characterize the Impact signals from the embedded sensors. A recurrent network is proposed for control of stress in the composite material. Preliminary results on a composite with embedded multimode fiber indicates that the speckle pattern of the optical fiber can be characterized for stress measurements. This new sensing technique will simplify the measurement components as compared to the conventional methods. 2.

Fiber optic Mach-Zehnder interferometric sensors for the measurement of various physical parameters such as pressure temperature and magnetic fields are well known and documented. This paper deals with the construction of a fiber optic sensor incorporating He-Ne and diode lasers. Preliminary experimental results for temperature sensing are reported.

We report the results of an experimental and modeling investigation of a high temperature superconducting opening switch. The switch was fabricated using a thin film of YBaCuO on a strontium titanate substrate and a Ar-Ion laser was used as a trigger source. When one of the critical parameters of a superconductor exceeds the critical value it makes a transition to a non-superconducting state (high resistance state). The change in resistance of a superconducting thin film due to electromagnetic radiation can be accounted for by two mechanisms:1 (1) the photon interacts with the thin film through optically induced phase process in the relative weak " boundary Josephson junctions" between the grains and (2) the radiation simply heats the thin film above the critical temperature (T). The dependence of the switching characteristics (switching time off resistance) on the bias (operating temperature and operating current) and trigger intensity will be discussed. Also the results of model based on heat equations and circuit laws will be discussed. 1.

We have developed two methods for thermal and mechanical stability analysis of optical mounts and components. The methods are capable of detecting movements on the order of 1 pm. We present a description and evaluation of the two methods along with analysis results of a 3-axis positioning stage. 1.

With the help of CODE V (a powerful tool for lens design) a fourteen element and two cemented doublet lens system with 0. 2 Reduction rate (Red) 10 cm field diameter 0. 32 numerical aperwre (NA) was obtained. The criteria for a microlithographic lens system applied to the lens will be discussed. 1.

It is recognized that unsymmetrical optical systems can provide better or unique solutions to certain optical design problems. However, the design of unsymmetrical optical systems is considered a difficult or obscure task. There are a large number of papers published dealing with isolated aspects of unsymmetrical optics; however, the number of papers presenting a theory and method for the systematic design of a particular class of unsymmetrical optics is small. There is today a significant understanding of the imaging properties of unsymmetrical optics, but there is not a general theory to design such systems as for example in the case of axially symmetrical optics. This is due to the great diversity of configurations in unsymmetrical optics. To design unsymmetrical optics, a number of design methods are available. In this paper four design methods that are based on theories still under development are briefly discussed. Ideally, a theory must provide insight about the main image characteristics or first-order properties, about image defects or aberrations, and provide a systematic method to conduct a design. An unsymmetrical optical system is defined as a combination of optical elements such as mirrors, lenses, and gratings, that lacks an axis of rotational symmetry.

A patented new injection-molding technology which already has been coninercially licensed nonexclusively to worldleading plastic magnetooptical (MO) mediamakers gives very high moldpacking pressures (needed for maximum microsurface replication) without high birefringence (minimized by reducing constrictiveness during fill) even with high viscosity plastics. I . " STATFMENT OF ThE PROBLIM" FOB ERASABLE OPTICALDI 5K MOLDINQ Al though gl ass and new cross 1 inkabl e p1 as t i cs are cont ending wi th polycarbonate and other injection moldable thermoplastics for this new market polycarbonate has prevailed as " first generation" erasable media substrate (although product life considerations may require added barrier coat to the MO coating stacks thereof). If performance of all these substrate contenders were equal the injection moldable thermoplastics will be preferred due to lowest cost and most automatable production If molded thermoplastics don''t ultimately win out a main reason will have been that these disk products require a long narrow mold cavity to be filled by the thermoplastic of choice then packed at double the pressures used for digital audio compact disks (CDs) in order to force the molten thermoplastic against the staner microsurface until detailed microreplication is assured yet only very low moldedin stresses can be tolerated (typically 1/2 to 1/5 those of audio CDs) 1. 1 Fillrelated Molding Problems M-O disks typically are 130am or 86nin in diameter and as molded thickness of a single

Coherent two-wave coupling in a photorefractive crystal (PRC) BaTiO3 is used to detect the presence of impurities (phase objects) in transparent liquids. Glass pieces of 100 microndiameter are easily detectable with a reliability of better than 96%.

3D networks and 2D network have been proposed for use in photonic switching systems and they may be able to provide higher density connectivity between logic devices than similar electronic implementations. In this paper optical implemen-. tations of both 2D and 3D crossover networks will be described. We show that crossover networks can be used for connecting multiple stages of 2-input 2-output switching elements as well as 2-input logic gates. We also show that these switching elements can be implemented using optical logic gates such as Symmeuic SEED devices. We then describe simple conversion steps that can be used to convert 2D crossover networks into 3D crossover networks. Since both network types can be implemented with low-loss optical imaging systems these networks may prove to be useful in future optical computing and photonic switching applications. 1. BACKGROIJND ON CROSSOVER NETWORKS The 2D crossover network is a multistage interconnection network that is based on crossover interconnectionsJ11 A 2D crossover network with N inputs and N outputs is shown in Fig. 1. The 2-input 2-output nodes within the network can be operated as 2-input 2-output switching elements or as 2-input logic gates. Crossover networks with 2-input 2-output nodes require that N where m is a positive integer. In addition fully-connected crossover networks have log2(N) nodestages with N nodes per node-stage and they have log2(N)-1 link-stages with N links per link-stage. In Fig. 1

We demonstrate a prototype digital free-space photonic switching fabric consisling of three cascaded 16x8 arrays of Symmetric Self Electro-optic Effect Devices used as logic gates. 1.

In this paper an overview of linear array cameras is given as well as techniques used to digitize the video signal. It begins with a description of charge-coupled devices (CCD) and linear array sensors. A brief survey of manufacturers of sensors cameras and digitizers is also included. Cameras discussed in detail are the Fairchild CAM13O1R Pulnix TJS000 Sierra Scientific 4050 TDI camera Reticon LC1902 and LC1932 color camera. Circular and color arrays made by Reticon are mentioned. Time Delay Integration sensors which can function with lower light levels than traditional linear array sensors are described. The signals used to control the operation of linescan cameras are listed and discussed. These include clock signals start of transfer signals and integrate signals. For example Reticon cameras/sensors require more than one clock signal. Since this paper involves connecting cameras to digitizers the RS-422 signal transmission standard is discussed and some integrated circuits that can be used to transmit and receive this type of signal are given such as 26LS31 and 26LS32 driver/receivers. Once we have a way of transmitting control signals and video signals we must connect the camera to a digitizer. If the digitizer doesn''t have RS-422 connections an interface is needed. An interface can simply be used to convert TTL to RS-422 or it can be a complex circuit producing and allowing variation of integration time and clock frequency. The interface may

When developing a low cost computed tomography (CT) imaging system using a PC compatible as a computer platform, it was necessary to create a data acquisition board for the PC with high enough precision (at least 12 bits per pixel) for CT imaging. Since the CT system uses in its standard configuration an image intensifier/camera detector with a rasterized video source (such as a TV camera), the data acquisition system (DAS) must be able to capture any part or all of a video frame. We created a DAS that is capable of 12-bit digitization at a speed of 10 million samples per second. Built on a single PC AT board, this DAS has all the functions necessary for video-speed single slice or volume CT data collection. The analog-to-digital conversion device (ADC) used is the Comlinea.r CLC925AI, which has 12-bit output. Since the DAS can also accept 16-bit digital input from an external ADC, the 12-bit samples are stored as 16-bit values. The DAS has two 1024 x 1024 x 16-bit image buffers and two direct memory access (DMA) paths, one to the AT bus and one to BIR's proprietary 64-bit data bus. The 64-bit bus connects the DAS, an image processing board, and a 32 MB memory board to provide the data transfer speed necessary to keep up with data acquisition and processing. It operates at a DMA transfer speed of 1 1.1 to 13.3 MHz. For transfers between the DAS buffer memory and the data bus, the DAS memory is a read-only data source; data cannot be written into the DAS buffer from the data bus.

One of the many promising features of free space photonic switching systems is the prospect of interconnecting many channels of high speed information. In these systems parallel sets of optical logic processors transfer information in the form of light that is imaged using conventional optical components. The use of imaging acquisition and processing systems is therefore naturally suited to the task of diagnostics in this situation. This paper will discuss image acquisition techniques that are being examined for photonic switching system diagnostics. 1.

In order to develop a computed tomography (CT) scanner that is substantially lower in cost than those currently available it was necessary to create an image processor that could meet the high CT data processing requirements while being compact and inexpensive. We designed and developed an image processing card that operates on the AT bus of a PC-compatible microcomputer. This card uses the Intel i860 RISC processor and custom hardware to allow simultaneous data collection correction and reconstruction. The design uses an architecture that allows separate image processors to work individually and in parallel so that additional processors can be added to speed up image reconstruction with no reprogramming. A single processor can reconstruct a 512 x 512 x 12-bit CT image in about 15 seconds a fully configured system can reduce the processing time to under 2 seconds. This processor is easily athptable to future CT data collection applications. 1.

Image restoration is a necessary step for higher level machine or human image analysis. In many applications image data are blurred and corrupted by additive noise. Because of the presence of singularities in the blurring operator regularization is an effective method for obtaining satisfactory solutions to image restoration problems. The application of regularization necessitates a choice of a regularizing parameter which trades fidelity to the data with smoothness of the restored image. For most problems of interest the choice of the regularizing parameter is not known a priori. Methods based on the properties of the residuals and on the generalized cross-validation have been proposed for estimating the regularizing parameter. In this paper alternative methods are proposed to compute the regularizing parameter. The resulting values of the regularizing parameter are compared with the values resulting from the above mentioned methods. Furthermore it is shown that under certain conditions all the above mentioned methods result in the same value for the regularizing parameter. Experimental results are presented which verify the previous theoretical results. 590 / SPIE Vol. 1396 Applications of Optical Engineering: Proceedings of OE/Midwest ''90

Real time image processing/computer vision systems require high performance parallel architectures in order to achieve required data processing speeds. In this paper a real-time architecture is proposed for a specific class of model-based vision algorithms that recognize objects in radar images [Morgan 89]. The architecture is simulated and the results of the simulations are related to currently available parallel processors in order to estimate the performance on these systems. Not only does this architecture solve the radar object recognition problem it can also be generalized for similar image processing/computer vision algorithms.

This paper examines several parallel image processing architectures designed for convolution filtering and solving least-square problems. Six papers on convolution are examined three describe linear systolic array designs and three describe special-purpose implementations. Four papers on median filter are examined. The algorithms used are all slightly different but the architectures are very different. Finally arrays that solve the least-squares problem using the QR decomposition method are examined. The criteria used to analyze the architectures are the intended environment the characteristics of the algorithms and the requirements of the host the cell and the array. 1.

This paper presents a new DOT coding scheme prioritized DOT (PDCT). This scheme uses partition priority coding 8 to perform PDCT. PPC encodes DOT coefficients in the order of their (absolute) magnitude in favor of the larger coefficients without coding overhead for position information. This results in an adaptive and progressive image coding. Different from classical DOT image coding this scheme does not use any bit allocations and quantization except integer rounding. Instead this scheme utilizes multiple distribution entropy coding (MDEO). This leads to an efficient coding for progressive image transmission or compression. For PDOT an image undergoes a 8x8 or 16x16 block DOT transform and the DOT coefficients are rounded to c-bit (typically 8 bit) precision. Using PPO the coefficients on the entire image are adaptively coded. That is the largest (i. e. the most significant) coefficients are coded first followed by the smaller ones until a given bit rate is reached. Using MDEO the values and position information (required by PPO) of the ordered coefficients are packed efficiently. The PDCT coding method is shown significantly better than current lossy compression methods in (1) SNR for a given bit rate (2) Adaptivity for images other than the " class" for which the method was originally tuned and (3) Early recognition of images when used in progressive image transmission. Typical SNR results are 30. 9 dB 34. 0 dB and 36.

The KCT system is designed for performing computed tomography (CT) imaging using the components of a real time radiography (RTR) system. The equipment necessary for CT imaging is provided as an RTR upgrade the components of KCT include a computer data acquisition system image processor and rotating mechanical stage. When integrated into an RTR system KCT provides the object scanning motion required for CT converts the analog video signal to digital data processes the data and displays the resulting cross-sectional CT images. The KCT computer is based on a PC compatible the data acquisition system and image processor are on AT boards in the computer. Because of its compact size inexpensive components and use of existing radiation source and detector KCT is available at a much lower cost than competing industrial CT systems. 1.

To assure quality in crystal production detection of the defects especially cracks and scratches is a vital task. Up to the present inspection operation on crystal blanks is still being carried out with naked eyes under an optical magnifier. Due to the tiresomeness or unconscientiousness which might result from the continual exposure to strong illumination the misclassification rate will unavoidably be high. Noting that the degree of the seriousness of cracks and scratches are different part of the defected crystal blanks (mainly those with scratches) could be retrieved to increase the productivity rate. In this paper a criterion is established to precisely defme the fme scratches and cracks based on the data collected from large amount of crystal blank samples which are known to contain cracks and/or scratches and algorithms are developed to automate the detection and defect-category identification. THE PROBLEM Automated inspection has been a very popular topic during the past twenty years not only in research and development but has already been put into practical use. A lot of successful experience has been obtained in modern factories in detecting the imperfection of the parts In many cases the computerized inspection coordinates very well with robotics and becomes an important link in the intergrated manufactured system. However in our crystal manufacturing industry such a development is completely new. As we know crystal products have widely been used. It has already

This paper presents an image processing algorithm which uses a centripetal self-adaptive line trace approach to locate the center of crn plants based on a detectable local feature of corn leaf. The experimental results show the usefulness of the algorithm for the machine vision applications related to corn plant identification and location. 1.

The necessary basis for any cost-effective highway and bridge deck pavement management program is reliable complete and up-to-date information concerning the state of pavement surface. An imaging system using the Shadow Moire method for the quantitative inspection of highway pavement surface in real-time has been developed which includes image analysis software for performing image enhancement and the analysis of Moire fringe patterns. Several enhancement schemes and thinning algorithms have been implemented and tested. The prototype system possesses a high degree of automation and gives enhanced pictures and accurate results. Also needed surface geometry information may be obtained efficiently and effectively.

A machine vision systelu has been developed to size and count ore as it passes down a conveyor belt. The imaging system consists of a line scan camera a zoom lens and a structured lighting arrangenient. The structured lighting produces a line of light which is projected onto the conveyor belt at an angle with respect to the caniera. When a piece of ore is present the line of light covering the piece is displaced out of the field of view of the camera. This method is used to overcome the poor contrast between the ore and the conveyor belt. Iniages are acquired using an iiaage processor which performs real-tiiae thresholding of the iraage before it is passed to a frame buffer. A coniputer progrant locates the ore pieces in the image and identifies the boundaries between those pieces which are touching. Chords corresponding to the approximate major and minor axes of each piece are calculated. Using these measurements the size distribution for a population of ore is determined. 1.

Imaging Science is rapidly emerging from its roots within the traditional disciplines largely as a result of recent advances in computer science and technology that have given rise to computer-based imaging devices (e. g. microscopes telescopes and medical imaging devices that produce digitized images) all of which can be understood in terms of a common set of principles and concepts. Specifically the development of very powerful high-resolution workstations high-speed networks mass data storage and software for dynamic interactive 3-D display and automated image analysis will have an increasingly profound impact on virtually every aspect of life and culture including scientific research and engineering health care and education as well as manufacturing commerce and industry. Advances in Imaging Science will be accelerated by the establishment of new synergistic relationships among " imaging scientists" who currently function in relative isolation within universities national laboratories and numerous industries. The UofC/ANL Center for Imaging Science has been established to foster such relationships. The State of illinois is in an ideal position to play a leadership role in the emergence of this important new discipline which is destined to have an increasing impact on the world economy. 1.

There are some unmistakable trends in the medical laser industry that have persisted throughout its short history. If we may assume that these trends will continue into the next decade, then we can deduce a few of the more obvious outcomes. However, as much as I would like to assure you of the reliability of these predictions, I must admit that what follows is merely science fiction for you medical laser buffs.

Some recent innovations in interferoinetric fiber optic sensors include special fibers new components and sensor systems. Many of the concepts have precedents in microwaves. 1. GENERAL PRINCIPLES The application of optical fibers to sensors is diffuse compared with their application to optical communications which is essentially focused on the single problem of how to get information from A to B. A fiber sensor is viable when it can do something not possible with better than more cheaply than any existing method. The probability of the emergence of a new sensor depends on the length of time that a need for the sensor and the possibility of meeting that need have co-existed regardless of whether the need or the possibility has appeared first. 2. TYPES OF SENSOR Fiber sensors can be divided into: a) Multimode fiber sensors which depend on amplitude effects b) Single mode (single path) fiber sensors which depend on phase effects. Since multimode fiber has existed for many decades the emergence of a new multimode sensor depends mostly on the discovery of a new need for such a sensor. On the other hand single mode/single path (i. e. polarization maintaining) fiber is relatively new and so is still being applied to existing needs. This is particularly so of recent innovations in fibers and components. SPIE Vol. 1396 Applications of Optical Engineering Proceedings of OE/Midwest ''90 / 709

Laser holography has been around since the early '60s('). The dazzling displays shown by Leith and Upatniek took the world by storm. Since then, due to fundamental limitations within the field and questionable business expectations of entrepreneurs, the pace of developing slowed down. Nevertheless, steady progress on both technical and business fronts are continuously made. In the '90s, we fully expect the fruition to take place both in the laboratory and the market place. New areas of applications are beginning to emerge, notably with respect to computers and computing, anti-counterfeiting, holographic optical elements (HOE), non-destructive testing (HNDT), multi-color display, high resolution (in space and time) imaging, quality control, and education. What follows is a brief discussion, in the above order, of each of these areas with some reasonable extrapolation of further developments during this decade.

Artificial neural networks are Information processing structures based upon models of brain functions. They are highly parallel and distributed dynamical systems that can carry out information processing by means of their state response to input stimuli. These networks promises advantages in speed adaptability and fault tolerance resulting from their highly parallel architectures. In this paper feedforward and recurrent networks are reviewed. These networks consist of collections of simple processing elements and dense interconnections among the processors. The optical techniques are good candidates for proving the neural network Implementation needs specifically the heavy interconnections among the processors. Finally some Important optical neural networks are reviewed.

Introduction of rewritable optical drives and disks during the late 1980's expanded the range of data storage applications for which optical recording is a preferred technology. The high density and reliability intrinsic in optical recording are combined with the ability to erase and rewrite data as in magnetic recording. Since it does not require flying a magnetic head in proximity to the media surface, rewritable optical media is removable and the risk of damage by a head crash or multiple uses is eliminated. Data can be written, read, erased, and rewritten millions of times without damage. Magneto optic (MO) recording is used in rewritable optical drives/media due to the advantages in speed and durability it offers relative to other methods (phase change, physical changes in polymers). MO drives and media are now available from several companies worldwide. 3M is a major supplier of media (disks) which are produced at it's Vadnais Heights, Minnesota facility. Each media cartridge offers 650MBytes of removable data storage on a two sided 130mm disk.

Single mode optical fiber systems are very attractive to the telecommunications industry because of low loss immunity to RF interference and high bandwidth. At present however commercially-available components are expensive bulky and have inconsistent performance. This paper describes a class of small size low-cost highperformance single mode 1 x n splitters optimally designed for the telecommunications industry. The high-silica 1 x n splitters are fabricated on a silicon wafer using a combination of flame hydrolysis deposition photolithography and reactive ion etching processes. This processing lends itself well to batch processing and is well controlled allowing for a high yield. We fabricated several types of single mode y-branch based 1 x n splitters using the above technology. By cascading simple 1 x 2 y-branch splitters 1 x 4 1 x 8 and 1 x 16 circuits were fabricated. For an unpackaged 1 x 8 splitter average insertion losses of 10. 0 dB including the y-branch splitting loss were obtained with a uniformity of 0. 3 dB. For a packaged pigtailed 1 x 8 splitter average insertion losses of 10. 4 dB with a uniformity of 0. 3 dB were obtained. 1.

With the emergence of fiber based links operating at 13 microns the need for high speed fiber compatible intensity modulators has increased significantly. We report on the development of a fiber pigtailed GaAs/AlGai. As based integrated optics Mach-Zehnder intensity modulator with speeds in excess of 3 GHz. 1.

The term binary optics was first employed to the best of my knowledge by Wilfred Veldkamp of MIT Lincoln Laboratory; although, these devices have been in existence prior to this of this term. Generally the term refers to optical components whose transmittance or reflectance occurs in binary steps. Usually fabricated by use of one or more binary masks. Other than simple diffraction gratings the earliest examples of binary optics would be computer generated holograms. Computer generated holograms, or more generally digital holograms, have been used since the middle 1960's, and like holography in general, digital holograms have found a few, but very few, useful applications. It has not been until thinking began to focus on more general diffractive optical concepts, and less on the principles of holography, that new applications began to materialize in the area of binary optics. Two exampies would be compact disc technology and the correction for chromatic aberation in refractive lenses.

This paper deals with the estimation of the motion field in digital angiographic sequences. An approach is developed according to which each frame is first segmented into a moving object of interest and the background. The original images are converted into binary images by using a Gaussian smoothing filter and thresholding. In the binary images pixels in moving objects have the value of " 1" and pixels in the background have the value of " 0" . The moving objects in the binary images are thinned to skeletons with unit width by using the Safe-Point Thinning Algorithm (SPTA) with restriction windows we introduce. Then a block-matching algorithm is used in estimating the motion for the pixels which belong to the skeleton. This approach to motion estimation results in reduced computations since only binary multiplications need to be performed for determining the match between two blocks. Therefore an effective searching method is proposed for finding the direction of displacement in successive skeleton frames. Very satisfactory results are obtained by applying the algorithm to 64 x 64. pixel digital angiographic image sequences. 1

This paper discusses components that are currently available for the physical layers of local area networks to be installed during the 1990s assuming that bit rates greater than 1 GBps-km are needed and that suitability for high-volume manufacturing is an important consideration. Components of this type include single mode fiber and laser diodes which have become widely used in longhaul telecommunications systems. Integrated optics and optical amplifiers offer additional capabilities in the areas of high-speed switching network couplers and taps and control of signal amplitude. 1

Optical system performance is dependent on what happens at the surface of optical elements and optical detectors. Optomechanics involves maintaining the correct shape and position of surfaces in an optical system. Alternately, optomechanics is sometimes defined as being involved with the mechanical engineering aspects of optical systems.

Gratings may be used to form images of objects illuminated with incoherent light. Under certain restrictions the object is either replicated in the image or a single filtered image is formed. This latter case has applications in microlithography. 1.

The use of reference mirror rotation during the production of a double-exposure hologram is discussed. The rotation of the mirror introduces a phase shift across the hologram that can be observed by scanning one''s head from one end of the hologram to the other. The sign ambiguity of the fringe order can be resolved by noting the apparent motion of the fringes as one scans one''s head across the hologram. This method is easily adapted to existing setups does not distort the original fringe pattern and the amount of phase shift across the hologram can be controlled easily. Sample experiments showing quantitative and qualitative results on a circular membrane are presented. 1.

In the recent past neural networks have been used in a variety of applications. The wide spectrum of applications has required invention of new architectures as well as improving existing architectures. This paper reviews some of different neural network architectures and their applications in optimization problems. The paper is divided into three primary sections: 1) a brief review of the type of problems that are suitable for neural networks 2) a brief description of some of the more commonly used neural network models and 3) a quick glance at application of Hopfield network to optimization problems. The first part will provide a quick glance at the two class of problems best suited for neural networks. The second part will briefly review various networks currently in use today. The application section will provide examples describing use of Hopfield network for solving economic load dispatch and linear programming problems. The paper also includes an extensive bibliography. 1.

An interferometer for aspheric testing ([AT) is under development at APA Optics for testing of general aspherics using inexpensive electron-beam written computer generated holograms (CGHs) as null compensators. This 152-mm aperture Twyman-Green interferometer is compatible with standard transmission spheres fringe analysis software and phase measuring accessories. Aspheric departures of up to several hundred waves can be measured using only standard interferometer accessories. Deeper aspheres may be tested using simple auxiliary optics. The interferometer configuration methods of operation and performance specifications are presented. 1.

Metal-Dielectric-Metal thin film assemblies have been used for more than forty years as absorption filters or reflection filters. Because this filter simultaneously absorbs nearly 100 of the incident radiation at certain wavelengths while reflecting nearly 100 at other wavelengths the filter is called either an absorption interference filter or a reflection interface filter. While the fundamental theory for understanding this filter is well established a combination of lumped circuit theory and distributed circuit theory offers a simpler approach to understanding how the filter works as well as offering a conceptual aid to designing the filter. These theories are combined in this paper to demonstrate this approach. 1.

We present a holographic interferometer using a photorefractive crystal SBN:60. A point source is recorded inside the crystal via two wave-mixing. By displacing the crystal laterally or radially a tunable Young''s interferometer or a tunable Michelson interferometer is constructed respectively. An advantage of this device is the elimination of errors due to hologram developing and repositioning processes. The interferometer is capable of producing sinusoidal gratings and zone plates. 1.

A new approach to realtime control of image contrast is described that is based on local scene statistics such as mean value and standarddeviation. The standardcontrastcontrolmethod observations ofthehighestandlowestpixelvalues within each set (Threshold detection) involves an inherently nonlinear feedback mechanism and is therefore difficult to control. A contrast controlloop using statistical detection techniques provides greater dynamic performance while decreasing sensitivity to scene anomalies such as sun glint when compared to contrast control loops using the traditional Threshold method. The Statistical approach is shown to offeradditionaladvantages incontrastenhancement when appliedto specificlightordarkregions of a scene. 2. THE CONTRAST CONTROL PROBLEM In order for the viewer of an image to recognize details of the scene and understand the picture the image mus have adequate contrast. By observing distantobjects on foggy days one sees that more distantobjects in areas oflowercontrast are very difficult to discern from neighboring objects at similar distances from the observer. As the distance between the observer and the objects decreases the features of the objects become clearer and more easily distinguished. Although the atmosphere is normally much clearer than the foggy example cited above the effects of the atmosphere on acquisition ofreconnaissance imagery is similar. Dust pollutants and other sources ofatmospheric haze contribute to reduce the contrast of the target scene on the ground as observed from the airborne platform. The effective reduction of scene contrast is particularly aproblem inlongrange oblique photography (LOROP)applications

Henry Draper described the use of flexible bands to minimize optical surface deflection of mirrors mounted in the optical axis horizontal position in 1864. Despite this historical background the theoretical basis for horizontal band support was not developed until quite recently. Use of finite element analysis has led to a better understanding of the performance of horizontal band supports. Analytical modeling of horizontal band supports has resulted in improvements in design. In particular the use of roller chains as a horizontal support significantly reduces the deflection of an axis horizontal large mirror. Providing that certain design precautions discussed in this paper are taken roller chains provide a relatively simple and low cost support. A number of successful roller chain supports for optics from 0. 68 m to 1 . 8 m in diameter have been built based on the results of the above analytical models. 2.