Integral photography, which provides continuous viewpoints and does not require any use of special glasses, is one of the most attractive methods for autostereoscopic display. To overcome the difficulties of pickup process and the pseudoscopic problem, a CGIP(Computer-Generated Integral Photography) method is desirable. In the CGIP scheme, the elemental images of imaginary objects are generated using computer instead of using pickup process. This easiness for generating elemental images makes the system compact and cost effective. For a wide viewing angle system, the f-number (the ratio of focal length to lens diameter) of the elemental lens should be small. However, the lens aberration is the penalty of the lens array with small f-number. Unlike the system using a conventional lens array, aberration occurs little in the case of adopting a well-designed Fresnel lens array which has small f-number. Fresnel lens array contributes to widen the viewing angle in the CGIP system. In addition, to increase the image depth range, dynamic integral photography is also proposed. The gap between a lens array and a display panel changes dynamically. Synchronized elemental image array for real and virtual mode is integrated in front of or behind the lens array. Therefore, observers get enhanced feeling of depth. In this way, autostereoscopic 3-D images with wide viewing angle and increased image depth can be obtained. The method can be usefully applicable to three-dimensional imaging.

We present an all optical photorefractive volume holographic processor for recognition of three-dimensional(3D) objects. The templates are recorded by use of a volume hologram in a photorefractive LiNbO₃:Fe crystal located at the Fresnel diffraction region and correlated in real time with a 3D object illuminated by coherent light. Experimental results for recognition of 3D objects are presented and compared with conventional two-dimensional method.

A new three-dimensional display system based on a volume-scanning method is demonstrated. To form a three-dimensional real image, an inclined two-dimensional image is rapidly moved with a mirror scanner while the cross-section patterns of a three-dimensional object are displayed sequentially. A vector-scan CRT display unit is used to obtain a high-resolution image. An optical scanning system is constructed with concave mirrors and a galvanometer mirror. It is confirmed that three-dimensional images, formed by the experimental system, satisfy all the criteria for human stereoscopic vision.

A color image is considered as a three-dimensional function in which the third dimension is the color axis. This approach permits the 3D generalization of the correlation between two images and the frequency filter design. In this work we analyze the properties of this 3D Fourier transform. Color pattern recognition is performed by means of a 3D correlation between an input colored image and the pattern to be detected. The spectrum whitening operation is analyzed in terms of the color transformation and some previously proposed element-wise transformations designed to improve the discrimination are revisited in this sense.

In this paper, a new multiview autostereoscopic display system using the volume holographic optical element (VHOE) is proposed. The VHOE is used for projecting the multiview images to the spatially different directions sequencially in time and it can be made from volume holographic recording materials. In this paper, Dupont photopolymer is used for VHOE materials and measured its physical properties such as sensitivity, diffraction efficiency for the optimal recording of diffraction gratings. In addition, from analyzing the dynamical aspects of optical characteristics for the photopolymer-based VHOE with coversheet, it is more useful for the practical implementation of the VHOE-based multiview 3D display system. By analyzing the basic and dynamic characteristics of Dupont photopolymer through some experiments, the feasibility for the implementation of VHOE-based multiview autostereoscopic display system is suggested.

Correlation filters are ideally suited for recognizing patterns in three-dimensional (3D) data. Whereas most model-based techniques tend to measure the overall dimensions of objects and their larger features, correlation filters can readily (and efficiently) exploit intricate surface details, the gray values of surfaces as well as internal structure, if any. Thus correlation filters may be the preferred approach in scenarios when intensity and range data are both available, or when the internal structure of an object has been mapped (e.g. tomography). In this paper, we outline the development of filters for 3D data that we refer to as Volume Correlation Filters (VCFs), illustrate their use with range images of an object, and outline future work for the development of 3D correlation techniques.

We propose a method of synthesizing computer-generated holograms of real-life three-dimensional (3-D) objects. An ordinary digital camera illuminated by incoherent white light records several projections of the 3-D object from different points of view. The recorded data are numerically processed to yield a two-dimensional complex function, which is then encoded as a computer-generated hologram. When this hologram is illuminated by a plane wave, a 3-D real image of the object is reconstructed.

In recent years, there have been many researches being done throughout the world on the 3D image conversion of 2D image. However, 3D image conversion of 2D image has many problems on obtaining the optimal stereopsis. Stereopsis is dominated to relative position of several objects and depth information within image. Accordingly, in this paper, as a new adaptive scheme for stereoscopic image conversion of 2D image is suggested. Two input images acquired by Stereo Camera have different disparity information to each other. Disparity map, based on disparity information, presents mutually different occulusion region in the left/right image. These depend on the left view & right view and front & rear view of the virtual image plane. If arbitrary threshold values are applied to disparity map, we can get segmented objects from the input image. Using the principle of horizontal parallax, segmented objects are shifted with optimal screen disparity. In this case, we can improve stereopsis by differential shifting.

Stereovision is an effective method in acquiring three-dimensional data from a real scene and forms a very active field of research. The conventional stereovision system usually consists of two or more cameras and therefore the system is bulky and has some problems like a camera calibration. In this paper, we propose a new scheme to acquire three-dimensional data from one camera and a lens array which consists of many elemental lenses. An algorithm that is appropriate for the proposed scheme is discussed and experimental results are presented.

Optical correlators using spatial light modulators in the filter plane have been discussed and presented for many years. In most cases, these devices are electrically addressed modulators and the filters are generated off-line in software. These filters can be tailored to fit the desired application but cannot be adjusted to account for real-time changes in the target appearance. In addition, building filters to account for all of the possible viewing configurations can strain the reference database and the memory storage capability of the system. Recently, a correlator architecture that used an optically addressed, multiple quantum well (MQW) spatial light modulator as the filter plane device was presented. The architecture is a modified Vander Lugt correlator with and additional input modulator. The filter formed by the interference of the reference image beam and the reference beam is recorded on the MQW spatial light modulator. The recorded filter retains the full complex information as high-resolution film did years ago. Additionally, the filter can be updated simply by changing the pattern on the reference input modulator. The second input modulator is used to address this stored filter in a normal correlator read-out configuration. The correlator has been completed and results will be presented as part of this paper.

Design of an on-board processor that enables recognition of a given road sign affected by different distortions is presented. The road sign recognition system is based on a nonlinear processor. Analysis of different filtering methods allows us to select the best techniques to overcome a variety of distortions. The proposed recognition system has been tested in real still images as well as in video sequences. Scenes were captured in real environments, with cluttered backgrounds and contain many distortions simultaneously. Recognition results for various images show that the processor is able to properly detect a given road sign even if it is varying in scale, slightly tilted or viewed under different angles. Recognition is also achieved when dealing with partially occluded road signs. In addition, the system is robust to illumination fluctuations.

We investigate the encoding of a phase-only filter together with a Fourier transforming lens using a single SLM. We consider the implementation of this composed filter with both phase-only and real-only SLMs. The restrictions on the optical setup parameters in terms of the SLM characteristics are discussed. We corroborate our proposals by presenting numerical simulations.

In this work we present three spatial filters that perform isotropic edge enhancement or recognition. Their design is based on first determining a convolution kernel that performs the desired operation and then, by Fourier transforming, obtaining the filter function. To the authors' best knowledge, these filters have not been proposed in the past. We present numerical simulations that corroborate our proposal. For the optical implementation of one of the proposed filters, a holographic technique capable of representing complex transmittances is required. In this case, the filter is simulated using a double-phase holographic code.

Fluorescent images of gallbladder stones, tissue and bile are obtained using a streak camera. A Match Spatial Filer (MSF) is made using a stone fluorescent image. The MSF is used to perform correlations with fluorescent tissue and bile image. A method for recognition of the stone and rejection of the tissue during the laser lithotripsy is proposed using the correlation outputs.

It is generally accepted that diffractive optical elements provide the optimal flexibility for the implementation of free-space interconnections between arrays of optoelectronic components. In this paper we discuss the Hopfield neural network demonstrator based around a 2D array of Vertical- Cavity Surface-Emitting Lasers and an off-the-shelf Si photodetector array. The interconnection matrix between the optoelectronic arrays is provided by a diffractive optical element, the precise nature of which is determined by the problem to be solved by the network. In order to study the potential of the optoelectronic Hopfield network topology for both scalability and miniaturization, the design and micro-fabrication techniques used in the creation of diffractive optical elements will be analyzed. This analysis will allow estimates to be made of the maximum number of devices that may be interconnected using diffractive optical elements as well as outlining possible approaches to system miniaturization by the use of micro- optical elements.

Pulse couple neural networks (PCNN) have demonstrated some very desirable properties. Chief among these is its ability to segment images very rapidly and very well. This capability has been demonstrated with many different types of imagery including synthetic aperture radar imagery, infrared imagery, optical correlator output imagery, and medical diagnostic imagery. Most of the implementations of this network have been done in software. Several attempts have been made to build electronic versions with varying degrees of success. Recently, an Army Phase II SBIR was awarded to incorporate a PCNN in a smart detector for both military and medical applications. One of the inherent difficulties in building an electronic PCNN is implementing the linking field that is the strength of this network. An optical implementation of the linking would potentially simplify the problem and take advantage of the inherent parallelism of optics. The resultant hardware could be simpler and faster than previous implementations making it an attractive solution. This paper will discuss the current status of the SBIR program, and present possible optical implementations using recently developed Vertical Cavity Surface Emitting Laser arrays.

In this paper we present a novel hardware architecture of a neural network system based on optoelectronic devices and electronic techniques. The main characteristics of the architecture are that it is fully interconnected, the interconnections are fully programmable, it avoids optical alignment problems, and is easily scalable to large numbers of pixel neurons. Description, experimental demonstration and discussion of the behavior of the architecture are presented.

In this paper, we provide a thorough analysis of decision boundaries of neural networks when they are used as a classifier. It has been shown that the classifying mechanism of the neural network can be divided into two parts: dimension expansion by hidden neurons and linear decision boundary formation by output neurons. In this paradigm, the input data is first warped into a higher dimensional space by the hidden neurons and the output neurons draw linear decision boundaries in the expanded space (hidden neuron space). We also note that the decision boundaries in the hidden neuron space are not completely independent. This dependency of decision boundaries is extended to multiclass problems, providing a valuable insight into formation of decision boundaries in the hidden neuron space. This analysis provides a new understanding of how neural networks construct complex decision boundaries and explains how different sets of weights may prove similar results.

A new imaging method which can obtain the gray levels directly from the output waveform of Pulsed Laser Radar (PLR) is developed. A simple digital signal processing technique and multi layer perceptrons (MLP) type neural network (NN) have been used to obtain the gray level information from the pulse shapes. The method has been implemented in a real PLR to improve contrast and speed of 2D imaging in PLR. To compare the method with the standard method, a picture consists of 16 gray levels (from 0 for black to 1 for white) with both method has been scanned. Because of the ability of NNs in extracting the information from nonlinear and noisy data and preprocessing of the noisy input pulse shapes to the NN, the average and maximum of errors in the gray levels in comparison with standard method more than 88.5% and 72.6% improved, respectively. Because in this method the effect of the noise is decreased, it is possible to make the imaging with the same resolution as in standard method but with a lower averaging in sampling unit and this dramatically increases speed of the measurements.

Computational complexity is the minimum price in terms of resources required to obtain the result using a given algorithm for a problem of a given size. Electronic computers can pay the price in terms of time or space - through concurrency methods such as parallel or pipelined processors. Optical computers allow the use of a third resource - fanin. By using space and fanin, optical processors can perform some tasks at a speed independent of the size of the input. Examples from conventional algorithmic processes and somewhat less familiar nonalgorithmic processes are provided as illustration.

We have developed an optical key distribution scheme where communicators are able to generate identical binary keys from a wideband optical phase-noise-bearing lightwave. In our scheme, a phase-noise-bearing lightwave is distributed among the communicators via optical fiber, and then converted an intensity modulation by an unequal path length interferometer and converted to a binary stream by a comparator. The use of a broadband noisewave (>100GHz, ~0.8nm at 1550 nm), precludes the possibility of an eavesdropper recording the signal in enough detail to analyze the noise in conjunction with the encrypted data processing. Corresponding intensity modulations are produced if both interferometers have the same pathlength inequality to within a certain tolerance. We have demonstrated identical key generation by two independent terminals receiving a distributed optical noisewave for bit positions designated with a good validity bit. A separate Ethernet link between the terminals was used to assess the quality of the binary key generation. The system uses three WDM optical channels to transmit the noisewave at 1550 nm, data at 1310 and a probe signal for interferometer stabilization at 1530nm.

The type of signals obtained has conditioned chaos analysis tools. Almost in every case, they have analogue characteristics. But in certain cases, a chaotic digital signal is obtained and theses signals need a different approach than conventional analogue ones. The main objective of this paper will be to present some possible approaches to the study of this signals and how information about their characteristics may be obtained in the more straightforward possible way. We have obtained digital chaotic signals from an Optical Logic Cell with some feedback between output and one of the possible control gates. This chaos has been reported in several papers and its characteristics have been employed as a possible method to secure communications and as a way to encryption. In both cases, the influence of some perturbation in the transmission medium gave problems both for the synchronization of chaotic generators at emitter and receiver and for the recovering of information data. A proposed way to analyze the presence of some perturbation is to study the noise contents of transmitted signal and to implement a way to eliminate it. In our present case, the digital signal will be converted to a multilevel one by grouping bits in packets of 8 bits and applying conventional methods of time-frequency analysis to them. The results give information about the change in signals characteristics and hence some information about the noise or perturbations present. Equivalent representations to the phase and to the Feigenbaum diagrams for digital signals are employed in this case.

This paper addresses a method to implement ternary filters with binary phase spatial light modulators (SLMs) for real-time optical pattern recognition applications using a filtering-based optical processor. A complex ternary filter can be considered as a binary phase function multiplied by a binary amplitude filter, which selects information by blocking or letting pass spatial frequencies. The main problem is that commercially available SLMs cannot provide amplitude and phase modulation at the same time and at high filtering rates. The method presented here involves coding the binary amplitude in phase only, by adding a linear phase to the frequencies representing undesired information. This information will be rejected outside the correlation plane, whereas useful information will pass without alteration. Computer simulations and experimental results have been obtained for different applications. These results show that the pure phase filters obtained are absolutely equivalent to the desired ternary filters and, furthermore, that any ternary filtering function can be easily implemented in a VanderLugt correlator that uses a binary phase SLM, thanks to this technique.

A digital holographic data reconstruction method with data compression is proposed. We show the number of quantization level of a digital hologram can be reduced. By computer simulations it is confirmed that the method is especially useful for binary images. For gray scale images, we propose a bit plane decomposition method. By this method, we show both high reconstructed image quality and a high compression ratio can be achieved. This method is applicable to both a normal digital hologram and an encrypted digital hologram.

At present, a topic of great interest for the scientific community is to obtain an automatic monitoring system of red tide blooming organisms. The advances in automated monitoring systems have demonstrated that this automation is possible. In this paper, an analysis of the problems in the automated identification of red tide phytoplankton blooming is presented, using an automatic optical-digital system. Specifically, interclass size differences of organisms, different rotation and localization of the organisms in a microscope field and interclass of these same properties. The analysis was done automatically using a liquid crystal display device as interface of the digital with the optical part. This analysis is done for the first time in a hybrid system in real time.

Tuberculosis (TB) and other mycobacteriosis are serious illnesses which control is mainly based on presumptive diagnosis. Besides of clinical suspicion, the diagnosis of mycobacteriosis must be done through genus specific smears of clinical specimens. However, these techniques lack of sensitivity and consequently clinicians must wait culture results as much as two months. Computer analysis of digital images from these smears could improve sensitivity of the test and, moreover, decrease workload of the micobacteriologist. Bacteria segmentation of particular species entails a complex process. Bacteria shape is not enough as a discriminant feature, because there are many species that share the same shape. Therefore the segmentation procedure requires to be improved using the color image information. In this paper we present two segmentation procedures based on fuzzy rules and phase-only correlation techniques respectively that will provide the basis of a future automatic particle' screening.

An agile optical scanning scheme is proposed that uses wavelength manipulations for deflecting a free-space optical beam by selection of the wavelength of the light incident on a wavelength dispersive optical element. Using fast tunable lasers or optical filters, this scanner features microsecond domain scan setting speeds, single/multiple beam(s) in space, and large several centimeters or more diameter apertures for sub-degree angular scans. The beam scanning scheme offers simple control (via wavelength tuning). The paper also introduces space multiplexing for optical beam scanning and discusses various system architectures utilizing both space and wavelength multiplexing to achieve high speed optical scanning with coarse and fine tuning capability. Experiments described demonstrate high-speed, high resolution, wavelength tuned optical scanning in one-dimension (1-D), two-dimensions (2-D), and three-dimensions (3-D).

We describe a logarithmic phase filter to extend the depth of field of incoherent optical hybrid imaging systems with a rectangular aperture. By introducing this filter at the system's exit pupil and digitally processing the detector's output, we were able to extend the depth of field by an order of magnitude more than the Hopkins defocus criterion.

In this work we propose a new wavelet transform based speckle denoising algorithm for SAR images. The algorithm will explicitly account for the signal dependent nature of the noise by studying the variances of detail wavelet coefficients. The algorithm will use the analysis of variance ANOVA technique to check if variances are due to means belonging to the same population or not. If neighboring variances indicate belonging to the same population, then it's a smooth region and coefficient should be smoothed. If neighboring variances indicate the presence of two different populations, then coefficient is due to image feature and should be preserved. This approach will provide the flexibility of adjusting to region intensity level and thus no need for the fixed threshold concept. The algorithm will take advantage of the fact that wavelet transform creates three detail sub-images and a coarse sub-image. Each detail sub-image is associated with frequency contents due to certain edge location and orientation. The algorithm will also consider using cross-information from all three-detail sub-images to decide whether coefficients are due to a feature and thus should be preserved, or they are due to noise and should be smoothed. Simulations will show that our algorithm will provide better performance in terms of PSNR, ENL , and visually than currently existing techniques.

The visual performance of liquid crystal displays (LCDs) has been usually inspected and evaluated by sensory analysis at the manufacturing process. One of the most indistinct visual problems is low-contrast non-uniform brightness region called muras. The accurate and consistent detection of the muras is extremely difficult because there are various shapes and sizes of muras and the inspection results tend to depend on the operators. We conducted a study on the quantitative evaluation of muras based on visual analysis and human perception. We converted the front of screen (FOS) images from the LCDs into distributions of luminance information, and the mura regions were distinguished from the background area using our novel algorithm. This approach also led to a weighting function for the categories of muras that appear in the panels. Our identification method can also distinguish between the muras caused by flaws in the LCD cells and the intentionally designed non- uniform luminance distribution of the backlight.

We design a two-way optical component network analyzer which can do the optical component two-way S-parameters measurement of fiber-optical signal processing element. Two kinds of basic fiber-optical signal processing elements are measured. The theoretical model based on Z-transform for deriving S-parameter are given. The experimental results can show the amplitude and phase response of S-parameter of optical signal processing elements. Such a two-way optical signal processing element S-parameter measurement technique can reduce human error in measuring the S-parameter of an optical signal processing element.

In this work, we use an optical amplifier with an active fiber loop mirror and a narrow-bandwidth fiber grating mirror to achieve an active fiber-optic FIR filter included IIR optical filter. The filtering characteristics of the active fiber-optic FIR filter included IIR optical filter are experimentally analyzed with considering the DC driving voltage of the active fiber loop mirror. The frequency response of the active fiber-optic FIR filter included IIR optical filter will be measured by a two-way optical component analyzer.

In this paper we present a brief review and examples of image registration. When one wishes to fuse multiple images from disparate sources, the first step is image registration, finding the appropriate transform which maps pixels in image 1 to the corresponding pixels in image 2. Only after this step one may begin to correctly fuse and process information from both images simultaneously.

One of the main challenges of any control (or image processing) paradigm is being able to handle complex systems under unforeseen uncertainties. A system may be called complex here if its dimension (order) is too high and its model (if available) is nonlinear, interconnected, and information on the system is uncertain such that classical techniques cannot easily handle the problem. Examples of complex systems are power networks, space robotic colonies, national air traffic control system, and integrated manufacturing plant, the Hubble Telescope, the International Space Station, etc. Soft computing, a consortia of methodologies such as fuzzy logic, neuro-computing, genetic algorithms and genetic programming, has proven to be powerful tools for adding autonomy and semi-autonomy to many complex systems. For such systems the size of soft computing control architecture will be nearly infinite. In this paper new paradigms using soft computing approaches are utilized to design autonomous controllers and image enhancers for a number of application areas. These applications are satellite array formations for synthetic aperture radar interferometry (InSAR) and enhancement of analog and digital images.

In this paper we present the results of applying data compression to a three-dimensional object recognition technique based on phase-shift digital holography. Industry-standard lossless data compression algorithms were first applied. Next, lossy techniques based on subsampling, discrete cosine transformation, and discrete Fourier transformation were examined. We used normalized cross-correlation in the object plane as our performance metric. For each hologram tested, we found that as many as 90% of the cosine and Fourier components could be removed, without significant loss in correlation performance.