This paper describes algorithmic development toward an automated process that generates a same patient sequence of fundic images that are normalized in position and intensity and have noise artifacts removed. Normalization of these fundic images is a key first step to further automatic analysis for the presence or progress of ophthalmic diseases. The Litton PRC team, inclusive of LItton Data Systems and Tomey Inc., evaluated the potential use of a hybrid optical/digital processor in the normalization of ophthalmic imagery. PRC provided the funding, initial architecture and approach for disease analysis using the optical processor. Tomey Inc. provided the fundus imagery and clinical advice on 'normalization' or images prior to undertaking the disease analysis. Litton Data Systems prototyped the algorithms and test using a rapid application prototyping tool for object recognition. The team approach was to do a coarse vessel alignment to bring the images within a to-be-determined level of alignment, and the repeat the algorithms at a threshold and alignment for fine vessels. The success of the coarse work encouraged the investigation of algorithms for processing higher resolution images with greater accuracy. The combined results completed the IRAD investigation and are shown as successfully aligning two images.
In this paper we investigate composite Fourier-plane nonlinear filters for distortion invariant pattern recognition. Different types of noise such as spatially non-overlapping color background noise, other non-target object, real scene noise, and additive white noise are used to investigate the performance of the filter. Computer simulation results are presented to show the performance enhancement using Fourier-plane nonlinearities in the composite filter designs. The training image sets and the test image sets are generate by in-plane and out-plane rotation of different aircraft.
Fractal image processing technology has been recognized as having great potential in automatic target recognition (ATR) and image compression. In this paper, Physical Optics Corporation demonstrates the feasibility of using a fractal image processing technique as a new and efficient approach for signature, pattern, and object recognition. Using optical Fourier transform and a ring-wedge detection technique, we generate and measure the power spectral density of an input scene. The log-log plot of the power spectral density vs. spatial frequency provides a very valuable signature for each input. Experimental results show that we can successfully discriminate man-made objects from natural objects in input scenes by analyzing signatures obtained this way.
We describe novel architecture for a real-time image restoration system of live TV signals. No DSP is involved. The spatial filtering is obtained from two electronic analog filters, one for the raster lines and one for the columns. The very fast response of analog filters is the key for truly real-time video frame rate performance. The digital part of the system serves the purpose of pipe-lined parallel data conversion and flow, but not that of image processing at all. Despite the lack of DSP, this architecture exhibits some very important advantages. It does not need any computational source, it is very fast, and it is much cheaper. Also our 'parallel analog computer' can be easily incorporated in any complex system with video signal data as a simple 'plug-in' between the camera and monitor. An important aspect is that the system carries lower digitalization noise than DSP, thus yielding better SNR characteristics at a lower price. The system is not bound to nay specific kind of spatial frequency filtering and can be electronically tuned to obtain exact performance parameters. Because of these advantages, this architecture is promising for a wide variety of system such as supermarket multicamera security, military and aerospace vision systems, and medical diagnostics.
A recursive formula expressing N-dimensional convolutions/correlations via (N-1)- dimensional ones is proposed. It can serve as a basis for implementing N-dimensional pattern recognition optically. Also a constant time optical algorithm for constructing N-dimensional convex hull and Voronoi diagram is proposed.
A filter function designed to maximize the peak-to-output-energy metric statically for non- overlapping target and overlapping scene noises is extended for use to detect and identify maritime targets within the framework of an automatic target classifier. The filter is designed to be distortion-invariant with respect to different target aspects. Computer simulations show that the filter can successfully detect the distorted true class targets. A set of tilers constructed using subsets of the true class input images were implemented to discriminate target aspects and shown to have reliable performance.
The goal of this work was to develop a fast optical correlator for automatic real-time target recognition. The tremendous importance of optical correlators for military and civilian applications was recognized recently and approved by a US conference committee of senators nd representatives. This publication presents the experimental results of detecting and identifying complex scattered signals by using an innovative, hybrid electro-optical correlator. Our technique is based on achieving optical correlation by utilizing state-of-the-art devices: time delay integration, charge coupled devices, liquid crystal displays, and electronically controllable light sources. Results of the experiment with our optical correlator, performed with simulated sonar signals with a center frequency of 100 kHz and duration of 8 to 512 pulses, show the possibility of recognizing a Doppler shift of 20 Hz. This Doppler shift corresponds to a target velocity of 20.7 m/sec. Simulation results indicate that we can achieve significant correlation for a noisy signal by using appropriate signal length. Our experiments demonstrate that we can perform approximately 1010 multiply accumulate operations per second with the high parallel optical corrector, compared to approximately 106 multiply accumulate operations per second using a Pentium 133 MHz personal computer. This new optical correlation scheme can provide solutions for overcoming the inherent shortcomings attributable to the low dynamic range of CCD, and the problem of compatibility caused by different pixel patterns between LCD and CCD by making use of high-quality optics and modern means of achieving uniform illumination.
Holographic recording techniques have been applied to implement synthetic estimation filters (SEF) for pose estimation of a docking spacecraft. Each SEF is a weighted combination of Fourier-plane hologram exposures of different target attitudes. These particular attitudes represent a span of the attitudes among which the SEFs will ratiometrically interpolate the attitude of the target in an input image. The pose is estimated by comparing the correlation peak intensities for all the filters with previously stored reference intensity peaks.
We analyze the performance of the nonlinear joint transform correlator in terms of output peak to peak standard deviation ratio. The main assumption used is that the signal energy is small relative to that of the additive noise; this assumption is defensible in practice due to the generally small spatial extent of target images relative to scenes. This work is an extension of that in two earlier papers. The analysis in these was in terms of the ratio of the expected values of correlation peak to noise floor. In the current paper we extend this to study the variability of the correlation peak itself.
Several designs for a reduced resolution optical correlator are proposed. The design of reduced resolution filters by multiresolution wavelet analysis (MWA) or by downsampling is extended to spatial light modular with fill-factors less than one. A reduced resolution optical correlator is constructed based on one of the above designs, and a comparison of MWA and downsampling filters is performed for different size targets. The experimental results show good qualitative agreement with simulation; however, the first-order correlation peaks were found to be greater for the experimental results. A possible reason for this is suggested and a new technique for measuring the fill-factor is proposed.
The leads on VLSI semiconductor flat pack chips are becoming more delicate and more closely spaced with time. Inspection is becoming more expensive because of the higher lead density and more necessary because of the increased risk of damage. A new optical system for lead inspection is proposed that uses self-interference of coherent light passing between the leads. According to the fractional Talbot effect, explained in the paper, there are planes of uniform intensity behind a perfect periodic structure such as an undamaged set of leads. Damage to leads takes a variety of forms. One or more leads may be skewed sideways with a potential for a subsequent short circuit to a neighboring lead after flow soldering. Alternatively, a lead may be bent up with a potential for an open connection on flow soldering. In our system, damage shows up as a deviation in intensity across the self-uniform plane. A computer simulation was written showing that small amounts of damage may be detected, measured and classified as skew or coplanarity or some combination. Laboratory experiments were performed to demonstrate the inspection capability and assess the difficulty in constructing a competitive system using this approach.
Volumetric memories based on the extension of conventional approaches by using the third dimension for multiple layers may offer dramatic increases in storage capacity. For such thick- media multilayer systems, the design of the optical head includes challenges in the control and tolerancing of aberration accumulation throughout the focal depth. In this paper we review the accumulation of these aberrations in high resolution thick disk media, and compare the tradeoffs associated with several aberration compensation techniques. Both analytical and simulation results of aberration accumulation as a function of the system f-number will be presented. Aberration compensation tradeoffs including compensator complexity, cost, compensation range, speed will b detailed. Simulation results of a compensation technique will be presented.
The recent and continuing construction of multi- and hyper-spectral imagers will provide detailed data cubes with information in both the spatial and spectral domain. This data shows great promise for remote sensing applications ranging from environmental and agricultural to national security interest. The reduction of this voluminous data to useful intermediate forms is necessary both for downlinking all those bits and for interpreting them. Smart on-board hardware is required, as well as sophisticated earth-bound processing. A segmented image is one kind of intermediate form which provides some measure of data compression. Traditional image segmentation algorithms treat pixels independently and cluster the pixels according only to their spectral information. This neglects the implicit spatial information that is available in the image. We will suggest a simple approach - a variant of the standard k-means algorithm - which uses both spatial and spectral properties of the image. The segmented image has the property that pixels which are spatially continuous are more likely to be in the same class than are random pairs of pixels. This property naturally comes at some cost in terms o of the compactness of the clusters in the spectral domain,but we have found that the spatial contiguity and spectral compactness properties are nearly 'orthogonal', which means that we can make considerable improvements in the one with minimal loss in the other.
Certain optical image processing schemes proposed over the years are severely limited in performance because optical intensity distributions produced by the associated processors are characterized by extremely low signal-to-bias ratios (SBR). The available dynamic range of standard image detectors is used up primarily by the bias content of the distributions, with little dynamic range left over for information-bearing signals. Examples arise in connection with incoherent holography, time-integration acousto-optical spectrum analysis and correlation, and OTF synthesis, as well as in other areas. What is needed for the processing of such low SBR image distributions is an imaging device that is essentially 'ac-coupled,' i.e., that rejects background bias components in low-contrast imagery and responds only to the signal components. The pyroelectric imaging camera, used traditionally for thermal imaging, has these characteristics if the bias components are non-time-varying and the signal components are made to change with time. Recent developments in pyroelectric materials and camera technology suggest that image processor outputs with SBRs of 10-3 or less can be detected with 8 to 14 bits of dynamic range going to the signal information alone. Thus, high SNR processing wit low SBR signals appears possible. We describe briefly the operation of pyroelectric detectors and cameras and then introduce a variety of real-time image processing applications.
In this paper, a new method to predict the classification error in the Gaussian ML classifier is proposed. The Gaussian ML classifier is one of the most widely classifiers in pattern classification and remote sensing because of its speed and performance. Several methods have been proposed to estimate error of the Gaussian ML classifier. In particular, the Bhattacharyya distance gives theoretical upper and lower bounds of the classification error. However, in many cases, the bounds ar not tight enough to be useful.In this paper, we proposed a different approach to predict error of the Gaussian ML classifier using the Bhattacharyya distance. We generate two classes with normal distribution and calculate the Bhattacharyya distance and the classification accuracy. The class statistics used to generate data are obtained form real remotely sensed data. We repeat the experiment about 100 million times with different class statistics and try to find the relationship between the classification error and the Bhattacharyya distance empirically. The range of the dimension of the generated data is from 1 to 17. From the experiments, we are able to obtain a formula that gives a much better error estimation of the Gaussian ML classifier. Apparently, it is possible to predict the classification error within 1-2 percent margin.
The recently developed technology of spectrally-adaptive light filtering has been applied to create a spectrally-matched filter directly in the optical channel, to select a target with a known spectral signature over an ambient background. The approach to ATR described eliminates intensive computer post-processing of multispectral images, making real-time target detection with enhanced signal-to-clutter ratio possible.
Transport of contaminants and bacteria in aqueous heterogeneous saturated porous systems have been studied experimentally using a novel fluorescent microscopic imaging technique. The approach involves color visualization and quantification of bacterium and contaminant distributions within a transparent porous column. By introducing stained bacteria and an organic dye as a contaminant into the column and illuminating the porous regions with a planar sheet of laser beam, contaminant and bacterial transport processes through the porous medium can be observed and measured microscopically. A computer controlled color CCD camera is used to record the fluorescent images as a function of time. These images are recorded by a frame accurate high resolution VCR and are then analyzed using a color image analysis code written in our laboratories. The color images are digitized this way and simultaneous concentration and velocity distributions of both contaminant and bacterium are evaluated as a function of time and pore characteristics. The approach provides a unique dynamic probe to observe these transport processes microscopically. These results are extremely valuable in in-situ bioremediation problems since microscopic particle-contaminant- bacterium interactions are the key to understanding and optimization of these processes.
Color pattern recognition based on projection preprocessing of red-green-blue components and single-channel matched filtering is described. We optimize parameters of projection preprocessing and correlation filters in terms of noise robustness. We design a phase-only filter optimized in sense of signal to noise ratio for optical pattern recognition. Computer simulation results are provided to illustrate color pattern recognition by using the proposed method.
The degree of protection for common types of security holograms is reviewed. The possibility of hologram recording directly from the holographic film being used for protection is demonstrated. We suggest the structure of security hologram allows one to prevent such unwanted copying. We also propose a detection system which provides for fast testing of a hologram's authenticity.
In this paper, we propose a new optical threshold generator as a key-stream generator for stream cipher systems. To implement the threshold generator optically, we use conventional twisted nematic type SLMs. This proposed system is based on the shadow casting technique for the AND operation between taps and register stages. It is also based on the proposed PMRS method for modular 2 addition. The proposed PMRS method uses the property of light's polarization on liquid crystal device (LCD) and can be implemented optically using one LCD and some mirrors. One of the major advantages of the proposed system is that there is no limitation of the number of the programmable tapping points. Therefore, this system can be used for high-security applications. Also, the proposed system can be applied for the 2D encryption system which processes large amounts of data such as 2D images. We verify the proposed system with some simulations.
This paper presents a new low-cost, multi-layer interconnection architecture, which has the potential of being utilized in a space environment. Instead of conventional electrical input/output (I/O) pins, optical I/O pins are used to provide multi-layer interconnection. A small vertical cavity surface emitting laser diode array and photodetector array are used to transmit and receive signals, respectively. A specially fabricated end-face lensed fiber array is utilized to provide high-speed, low-loss, and low-crosstalk signal transmission. By employing this new optical I/O pin architecture, many types of standard multi-layer interconnection can be easily realized. This presentation covers the main concepts, major components, fabrication procedures, and expected applications of this architecture.
An all-optical reconfigurable crossbar switch is implemented by unique integration of dynamic photopolymer technology, simple geometry optics, and waveguide technology. This all-optical crossbar switch provides high reconfigurability, large fanout, and radiation hardening. These characteristics are essential for both parallel/distributed processing and high-bandwidth communication, especially for B-ISDN.
In this paper, a substrate-guided-wave-based perfect shuffle (PS) having an 8-to-8 interconnection is demonstrated at 632.8 nm wavelength. Sixteen waveguide holograms are fabricated on a planar waveguiding plate with the full functionality of the PS. The diffraction efficiencies of the waveguide hologram are all within 80 percent +/- 5 percent. The planar architecture combined with the surface-normal configuration of the demonstrated PS makes the integration with vertical cavity surface emitting lasers and other surface mountable processing elements highly feasible.
An optical communication/switching system with data speeds well beyond any state-of-the-art electronic board-to-board computer interconnect is proposed. This interconnect is based on the use of microprism arrays to achieve large-scale, wideband interconnections for optoelectronic systems.
We constructed an optical system for performing 2D logic operations. We demonstrated the logic operations of OR, NOR, XOR, AND. The photoinduced dichroism of Bacteriorhodopsin (bR) is the physical mechanism exploited in the optical system. In its normal state with no light illumination a bR doped polymer film is isotropic with random distribution of bR molecules. When excited by linearly polarized light, only those bR molecules whose transition dipole moments for absorption lie in or near the direction of the electric field ar bleached at the actinic light wavelength due to the B to M transition. If a probe beam is incident in the regions illuminated by the actinic light it will no longer be interacting with an isotropic film. The film is now anisotropic and dichroic. Due to dichroism the actinic light illumination produces an angular rotation of the plane of polarization of the probe beam. In the experiments the bR film is kept between two crossed polarizers to get zero output from the probe beam at a screen when no actinic light is incident on the bR film. The two polarizers and the screen are kept in the path of the probe beam. There are two actinic light beams that induce dichroism and they are made orthogonally polarized with respect to each other by the use of a polarization rotator. The polarization rotator can be arranged so that their polarizations are made parallel for some of the logic operations. The plane of polarization of these two actinic beams are at 45 degrees to the pane of polarization of the probe beam. The combination of rotation of the analyzer and the polarization rotation of one of the actinic beams allows for many logic operations to be performed. No interference recordings are involved in the experiments and hence vibration isolation systems are not required. A coherent source is not a requirement either since a white light source with an appropriate wavelength filter can induce photoanisotropy in the bR film.
The application of volume M-type holograms for building multichannel geometries in pattern recognition system is considered. The results of theoretical and experimental investigations of the hologram's parameters as a function of their recording parameters and their use as filters in correlation setups are presented. Multichannel correlation schemes where the processed signals have either different or the same wavelength are proposed. We have shown that all the correlation schemes proposed allow one to increase data throughput several times over single- channel BR-based correlators.
This paper describes novel multiple quantum well optical modulators in which the optical modulation is achieved by field dependent birefringence in strained MQW layers. The modulator operates under normal incidence. In our study, field dependent birefringence is obtained by straining the MQW layers using Surface Acoustic Waves. Computations of refractive index change, absorption coefficient and contrast ratio are presented. The devices are being fabricated using lattice matched InGaAsP/InGaAsP layers on InP substrates. The polarization-sensitive birefringent spatial light modulators are different from conventional multiple quantum well SLMs and self electro-optical devices which generally utilize electro- absorption.
The search for a dynamic recording medium that can be used in real time without the need for processing has become a critical issue in the development of practical neural network systems, correlators, all-optic switches, image and signal processors, and optical storage devices. A typical optical material respond to changes in the intensity, polarization, or wavelength of the illuminating light. The optical material developed and used for neural network applications responds to changes in the polarization of blur or green laser light. Implementing a neural network or performing optically-controlled acoustic beam steering requires a high-speed read/write/erase optical memory. POCs erasable dye polymer material offers a high read/write/erase data rate, nondestructive reading, fast data access, high storage density, overwrite capability, and long cycle life.
A novel compact planar configuration for correcting the asymmetric divergence of light emanating from diode lasers is presented. It is comprised of two holographic lenses that are recorded on one transparent substrate, where the light propagates form one lens to the other by means of total internal reflections. The design of the overall planar configuration is presented along with experimental result. The results reveal that it is possible to focus a collimated asymmetric beam to a circular spot.
Some regular fractals, as Cantor bars and Sierpinski carpet, can be obtained as multiplicative superposition of periodical functions. Adding an exponent to each of this functions we can obtain a system to apply in optics for image processing, because different combinations can be achieved. A parameter to characterize the fractal structures in the Fresnel and Fraunhofer regions is introduced. It is called the in-order self-similarity function, which permit us to determine the periodical components filtered from the initial structure. The application is developed mainly for 2D fractals as the Sierpinski carpet.
Two hybrid optical elements re presented for dual-focus pickup in the digital versatile disk (DVD) system. One is a hybrid dual-focus lens and the other is a holographic regional mirror. The optical head consisting of these two hybrid optical elements can read the information recorded on two layers of DVD simultaneously. Its specifications are described. The fabrication parameters of these two hybrid optical elements are derived.
Two cryptographic methods for storing graphic information are presented in this paper. The information is encoded by phase modulation or the angular coding. The information are read out by means of holographic interference. Without decoding card, no one can read out the information stored in the informational card.
In this paper, an optical processor for real-time human face recognition is presented. This processor is based on the method of modified complementary encoding hit-miss morphological transform, which combines the foreground and background of an image into a whole. The processor makes use of an incoherent correlator for optical implementation of hit-miss transform in one step and only one commercially available liquid crystal display panel as two real-time SLMs for both input face image and matched face image. The experiment result have shown that the processor has a recognition speed of 10 frames per second, accuracy over 90 percent and error tolerance to distortion-variance within 8 degrees, noise disturbance up to 25 percent and image losing up to 50 percent. The processor has potential applications on real- time human face recognition in customs, banks and stores for security.
Pattern recognition using a coherent optical correlator has many advantages, including high speed operation at almost the speed of light and the implementation of parallel processing. The key part of an optical correlator is the so-called optical filter. Properly designed filter should clearly indicate the presence of desired features in an image to be detected. Therefore,the performance of an optical system relies essentially upon the performance of the filter. Much research have been conducted to improve the performance of optical correlators. Most approaches to the filter design, however, fall short of providing robustness to minute changes in the image. In this paper a new approach to the adaptive design of optical filters, which are both shift- and scale-invariant, is proposed. The filters are constructed in real-time in an optical pattern recognition system by an adaptive, iterative numerical approach. The design is formalized as an optimization procedure, for which the filter performance is the function to be maximized. During the training procedure filter parameters are selected to maximize the distinction between the target and other objects in the image. The latter problem is solved using the genetic algorithm. Filters obtained in this optimization procedure are good discriminators since they utilize all the visual information about the target. Computer simulations demonstrate high discrimination of the designed filters.