Recently with the development of pickup and display devices, real-time integral photography or integral imaging has been an attractive method over other techniques for displaying three-dimensional images. However, previous methods basically belong to the transmission-type display. In this paper, a reflection-type integral imaging is studied and some experimental results are shown to prove its feasibility. This can be implemented by adopting a concave mirror array instead of a convex lens array. A concave mirror array in the reflection-type functions like a lens array in the conventional transmission-type. Thus, the 3D image is integrated by the rays that reflect from a mirror array. We could obtain autostereoscopic images with full parallax. In addition, if we combine the reflection-type system with another conventional transmission-type system, the feeling of depth can be more enhanced. A schematic of a concave mirror array on a curved surface is also proposed and discussed. The curved surface concentrates the rays to a small-sized display panel such as a beam projector. We expect this new form of integral imaging will contribute to realize a practical 3-D display system.

We propose the use of synchronously moving micro-optics (lenslet arrays) for image pickup and display in three-dimensional integral imaging to overcome the upper resolution limit imposed by the Nyquist sampling theorem. With the proposed technique, we present an all-optical three-dimensional integral imaging projector. An optically addressed spatial light modulator is used, which potentially provides better image resolution than the conventional CCD and liquid crystal display pair. We present experimental results using a liquid crystal light valve.

The Imaging Technologies group at De Montfort University has developed an integral 3D imaging system, which is seen as the most likely vehicle for 3D television avoiding psychological effects. To create real fascinating three-dimensional television programs, a virtual studio that performs the task of generating, editing and integrating the 3D contents involving virtual and real scenes is required. The paper presents, for the first time, the procedures, factors and methods of integrating computer-generated virtual scenes with real objects captured using the 3D integral imaging camera system. The method of computer generation of 3D integral images, where the lens array is modelled instead of the physical camera is described. In the model each micro-lens that captures different elemental images of the virtual scene is treated as an extended pinhole camera. An integration process named integrated rendering is illustrated. Detailed discussion and deep investigation are focused on depth extraction from captured integral 3D images. The depth calculation method from the disparity and the multiple baseline method that is used to improve the precision of depth estimation are also presented. The concept of colour SSD and its further improvement in the precision is proposed and verified.

A novel optical correlator for three-dimensional (3-D) object recognition is proposed herein. Several projections of a 3-D scene are recorded under white light illumination and fused into a single complex two-dimensional function. After properly filtering this function, it is then coded into a computer-generated hologram (CGH). When the CGH is coherently illuminated, a correlation space between the 3-D tested scene and the reference function is reconstructed, in which light peaks indicate on the existences and locations of true targets in the observed 3-D scene. Experimental results are presented.

Integral imaging has been received much interest due to its various advantages. However, the narrow viewing angle is one of the bottlenecks of this technique. Basically, the viewing angle is limited because the displayed area of elemental images is confined. In addition, the interference of neighboring elemental lenses causes the integrated image overlapping. Recently the viewing-angle-enhanced scheme with lens switching has been proposed. In that scheme, a mechanical moving part with fast speed to obtain a natural after-image was required. In this paper, we propose and demonstrate a novel scheme of integral imaging using a mask with two devices to enhance the viewing angle. This is implemented by two identical systems that consist of a lens array and a display panel. The images from the two sets of devices are combined with a beam splitter. A mask that has alternately repeated on/off patterns in an array form is attached to each lens array and the area of on/off in two masks are opposite (interleaved). The integrated images in each system are combined spatially and the viewing area is doubled. Thus, it can be stated that spatial multiplexed images instead of the time-multiplexed images contribute to enhance the viewing angle of integral imaging without any mechanical movement.

Several methods of concealing an image in a different hardcopy image are presented. The hidden image is secured since only a spatial digital, or optical, correlator equipped with an undecipherable key function can reveal the concealed image from the visible picture. The techniques and their robustness to noise and distortions are demonstrated.

We describe a unique approach to image resolution enhancement, inverse kriging (IK), which takes advantage of the spatial relationship between high- and low-resolution images within an area of overlap. Once established, this mathematical relationship then can be applied across the entire low-resolution image to significantly sharpen the image. The mathematical relationship uses the spatial correlations within the low-resolution image and between the low and high spatial-resolution imagery. Two of the most important requirements of the technique are that the images be co-registered well within the resolution of the larger pixels and that the spatial structure of the training area (where the spatial correlation statistics are computed) is similar to the structure of the remaining image area where it will be applied. Testing was performed using same-sensor and multi-sensor imagery. We show results that indicate that the method does improve the low spatial-resolution imagery. The selection of a training area spatial structure similar to the area being processed is important, as areas with different spatial structure (e.g., vegetation versus buildings and roads) will produce poor results. Comparisons with bilinear interpolation demonstrate that IK could be used as an improved interpolation tool, for example, in the image-registration process.

The technique of the multiple phases retrieval algorithm (MPRA) for designing optical security and verification systems is proposed in this paper. This technique is based on a 4-f optical correlator, which is a common architecture for optical image encryption and verification systems. In the proposed systems, however, two or more phase masks are iteratively retrieved by using the MPRA to obtain the target image. The convergent speed of the iteration process in the MPRA is significantly increased and the recovered image is much more similar to the target one than those in previous approaches. Moreover, the system security is increased since only the pair-wise retrieved phase masks can correctly recover the target images. To avoid carrying two phase keys, one of the phase mask serves as the key and the other phase mask can be stored in the database of the security system as an active lock. Finally, according to our simulation results, the misalignment effects for the phase mask in the Fourier plane are more series than that in the input plane.

In this work we describe the framework of a functional system for processing and analyzing images of the human eye acquired by the Hartmann-Shack technique (HS), in order to extract information to formulate a diagnosis of eye refractive errors (astigmatism, hypermetropia and myopia). The analysis is to be carried out using an Artificial Intelligence system based on Neural Nets, Fuzzy Logic and Classifier Combination. The major goal is to establish the basis of a new technology to effectively measure ocular refractive errors that is based on methods alternative those adopted in current patented systems. Moreover, analysis of images acquired with the Hartmann-Shack technique may enable the extraction of additional information on the health of an eye under exam from the same image used to detect refraction errors.

Most engineered products/processes/systems have continually evolved to enhance their performance. However, they still need to further evolve towards having such characteristics as high precision, intelligence and autonomy. In this paper, an opto-mechatronic systems technology is introduced as one of the enabling techniques, for such evolution. It is an integrated multidisciplinary technology combining optical, electrical, mechanical and computer engineering fields, thus creating new value and functions as a result of integration. Such integrated technology is shown to be derivable from the key technologies of the optical and mechatronic engineering fields and to a variety of fundamental functionalities required to produce the systems. The obtainable synergy and the future perspectives of the technology are discussed in detail.

The inverse computation for the design of phase masks in two layers is investigated and a novel algorithm presented. Phase masks are also referred to as phase-only holograms or diffractive optic elements (DOE). A two layer system can transform an arbitrary specified light field at an input plane to a desired light field at an output plane. The light field includes both intensity and phase. Such a system can be cascaded for higher level functionality. There are two computations involved. The first computes a sensitivity matrix symbolically. The elements of the matrix hold the variation in each element at the output plane with variation in each element of both phase screens. An element of this matrix is provided for reference. The second algorithm iteratively updates the phase screen values to bring the output field to that desired. On each iteration, the algorithm performs a forward computation from input to output. The phase values are updated using the sensitivity matrix and the error at the output relative to that desired. A unique solution is not required, only one that provides the required transformation from input plane to output plane.

In this paper we present an optoelectronic hardware implementation of a neural network system based on optoelectronic devices and electronic techniques. The system is composed of basic cells with optoelectronic artificial neurons; every basic cells exploits the communication strengths of optics to broadcast the input to all neurons and the computational strengths of electronics to assign the interconnections weights. Description of the architecture of the basic cell, the first implementation of a prototype based on the proposed system and examples of configuration of the neural system are described.

Based on the Lau Effect we propose a system suitable for eliminating non-periodic noise on bi-dimensional periodic structures in which a liquid crystal display (LCD) is used as a dynamic grating source. Experimental results obtained with our proposal are shown.

It has been shown in the extensive literature that wavelets are applicable to data processing. However, two shortcomings exist in using this mathematical technique for real-time in-line data processing. These must be addressed before any robust wavelet processing architectures can be created and applied to data in a non man-in-the-loop fashion. The first is autonomous selection of the basis function, and the other knowing when to stop acquiring data and invoke the wavelet transformation. Once these two ambiguities are resolved, autonomous feature selection algorithms can be created and the utility and performance of the resulting wavelet features evaluated. In January 2002, the authors began an IRAD study to examine varioius proposed methods for resolving these issues. This paper presents the results to date for generating wavelet transform basis functions from given 1-D time series data.

In this paper, we describe a new method of processing LADAR imagery for detecting and recognizing targets. LADAR is a unique sensor in that it allows data to be manipulated in 3D coordinates and "range-corrected" so that all objects are represented at their true size. For correlation-based detection methods, this leads to a significant reduction in the number of filters that need to be used. The detected regions of interest can be processed in 3D space using distortion tolerant volume correlation filters. We also discuss the performance of the algorithms for target detection on a set of LADAR images.

A Kalman Filter based algorithm capable of detecting and discriminating one or more aerosol clouds as a function of range will be presented. The traditional Differential Scattering (DISC) technique does not optimally utilize all the information available with tunable LIDAR sensors. For this reason, the authors have investigated an alternative approach that can better handle the general multi-material multi-wavelength scenario. The processing is developed around a statistical signal model that includes additive noise and the effect of a finite laser pulse duration. This algorithm was tested using data that was generated to simulate the response of the Army's FAL sensor. The algorithm is shown to be able discriminate between three materials.

Most organizations usually have large archives of paper documents that they maintain. These archives typically contain valuable information and data, which are imaged to provide electronic access. However, once a document is either printed or imaged, these organizations had no efficient method of retrieving information from these documents. The only methods available to retrieve information from them were to either manually read them or to convert them to ASCII text using optical character recognition (OCR). For most of the archives with large numbers of documents, these methods are problematic. Manual searches are not feasible. OCR, on the other hand, can be CPU intensive and prone to error. In addition, for many foreign languages, OCR engines do not exist. By contrast, our system provides an innovative approach to the problem of retrieving information from imaged document archives utilizing a client/server architecture. Since its beginning in 1999, we have made significant advances in the development of a system that employs optical correlation (OC) technology (either software or hardware) to access directly the textual and graphic information contained in imaged paper documents therefore eliminating the OCR process. It provides a fast, accurate means of accessing this information directly from multilingual documents. In addition, our system can also rapidly and accurately detect the presence of duplicate documents within an archive using optical correlation techniques. In this paper, we describe the present system and selected examples of its capabilities. We also present some performance results (accuracy, speed, etc.) against test document sets.

In this paper we present a comparison between four invariant face recognition techniques: synthetic discriminant function (SDF) based recognition, projection-slice SDF based recognition, optoelectronic correlator based neural network, and pose estimation based recognition. The pose estimation technique does not involve composite image generation and is most successful of these techniques mainly due to its success in maintaining a low interclass crosscorrelation. The optoelectronic fringe-adjusted joint transform correlator (JTC) is used to provide correlation. A description of the optoelectronic system and the simulation results are provided.

Fractional correlation is an extension of the conventional correlation. It employs fractional Fourier transform (FRFT) that includes the conventional Fourier transform as a special case where the order of the FRFT equals one. Because of the FRFT's lack of the shift-invariant property, the FRFT is not applicable to the conventional joint transform correlator, but to the nonconventional joint transform correlator (NJTC) that have been proposed by F. T. S. Yu et al., in which separate lenses transform the input signals and their spectral distributions overlap on the square-law detector. This provides an optical implementation of the fractional correlation. The conventional Fourier transform generally yields a high peak at the center of the spectral plane. But the FRFT gives a spectral distribution with no high peak, which is desirable because the square-law detector has a finite dynamic range for the linearity. Moreover, we prove that the fractional correlation produces a narrower output distribution and has the same correlation value at the center of the output plane as the conventional correlation. The conventional correlation has the shift-invariant property, but the fractional correlation has not.

Problems associated with nonlinear pulse and beam propagation, especially those involving higher-order terms in the nonlinear Schroedinger equation, usually require robust numerical techniques for their solution. In this paper we utilize an adaptive wavelet transform in order to investigate optical pulse self-steepening and optical beam self-focusing, as well as higher-order nonlinear terms which cannot be approximated by higher-order derivatives. Additionally, we show that the numerical method developed herein can be used to approximate any order of nonlinearity in the self-phase modulation term. The adaptive capability of the discrete wavelet transform developed herein allows this technique to accurately track the steep pulse gradients associated with higher-order terms by adaptively switching to higher, more accurate wavelet levels. By utilizing this adaptive wavelet transform technique, one can perform analysis of any terms in the NLS equation entirely in the wavelet domain without the need for resorting to a split-step method, as is often the case.

Integral imaging (or called integral photography) is an attractive three-dimensional display method because of its many advantages over other three-dimensional display methods. However, the thickness of the displayed three-dimensional image which can be expressed is limited by various optical parameters of the system and is relatively small. In this paper, we propose a method to increase the thickness of the displayed three-dimensional image without severe resolution degradation by adopting a birefringent material and a dynamic polarizer. We explain the principle of the proposed method and verify it experimentally.

Optical inspection techniques have been widely adopted in industrial areas since they provide fast and accurate information on product quality, process status, and machine conditions. The technologies include sensing using vision, laser scattering and imaging, x-ray imaging, and other optical sensing, and data processing for classification and recognition problems. Frequently, data processing tasks are very difficult, which is mainly due to the large volume, the complexity, and the noise of the raw data acquired. Artificial neural networks have been proven to be an effective means to cope with the problems difficult to solve or inefficient to solve by convectional methodologies. This paper presents the applications of neural networks in optical inspection tasks. Among the variety of industrial areas, this paper focuses on the inspection tasks involved in printed circuit board manufacturing processes and semiconductor manufacturing processes, which are the most competing industries in the world today. In this paper, the inspection problems are addressed and the optical techniques together with neural networks to solve such problems are reviewed. The application cases to which neural networks are applied are also presented with their effects.

In this paper we demonstrate a novel phase contrast system that employs a BR film. Since the filter is optically induced by the Fourier transform of the phase object, no alignment is necessary at the filter plane making it extremely robust. Due to the optical properties of BR films the phase filter can be induced with low light intensity levels. The material response allows operation at video frame rates, processing of high spatial resolution objects, and the use of relatively inexpensive laser sources. Such characteristics and the fact that BR films can be produced at a low cost makes the system simple to implement, relatively inexpensive and extremely robust. The effects of varying the illuminating area beyond the phase object area and filter saturation are also analyzed.

In this work we propose a convolution kernel capable of realizing phase visualization when operated with a phase function. The proposed convolution kernel is a binary complex function. We present numerical simulations of its performance on one and two-dimensional binary phase functions. We also briefly discuss its implementation in a joint transform architecture and the requirements imposed by the detector at the system's Fourier plane. Finally, we analyze the effect of quantizing the Fourier data on the contrasted output images.

Absolute optical angle sensor is described that is an essentially digital opto-electronic device. Its purpose is to resolve the relative and absolute angle position of coded disk using Pseudo-random-code and Triangle-code. In this technique, the angular position of disk is determined in "coarse" sense first by pseudo-random code. A further "fine" angular position data based on pixel count is obtained by Triangle-code which result 0.006° resolution of the system provided that 7 μm line image sensor are used. The proposed technique is novel in a number of aspects, such that it has the non-contact reflective nature, high resolution of the system, relatively simple code pattern, and inherent digital nature of the sensor. And what is more the system can be easily modified to torque sensor by applying two coded disks in a manner that observe the difference in absolute angular displacement. The digital opto-electronic nature of the proposed sensor, along with its reporting of both torque and angle, makes the system ideal for use in intelligent vehicle systems. In this communication, we propose a technique that utilizes pseudo-random-code and triangle-code to determine accurate angular position of coded disk. We present the experimental results to demonstrate the validity of the idea.

An analytical approach using human perception has been introduced to evaluate the front-of-screen (FOS) quality of liquid crystal displays (LCDs), in particular regarding the regions of the liminance non-uniformity called "mura". The word "mura" is a Japanese term similar to "blemish" and has been adopted in English to provide a precise name for certain imperfections of the display pixel matrix surfaces that are visible when the display is in active use. The accurate and consistent detection of the mura is extremely difficult because there are various shapes and sizes of mura and the inspection results tend to depend on the inspectors during the LDC manufacturing process. We conducted a study on the quantitative evaluation of mura based on visual analysis, intending to clarify the detection method and create an automated mura inspection process. We developed an algorithm to extract and evaluate mura by using a hardware system based on a commercially available CCD camera and a PC with an image processor board. This system can successfully identify and evaluate mura. 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. In order to further match the evaluation results of mura to human perceptions, we conducted a perception test with some expert inspectors by using pseudo mura and this approach led to categorizing "just noticeable differences" according to the varieties of mura. This paper describes the research in human perception and the approach adapting the intrinsic rules of sensory analysis to the quantitative evaluation of mura.

A new fast algorithm of 2D DWT transform is presented. The algorithm operates on byte represented images and performs image transformation with the Cohen-Daubechies-Feauveau wavelet of the second order. It uses the lifting scheme for the calculations. The proposed algorithm is based on the "checkerboard" computation scheme for non-separable 2D wavelet. The problem of data extension near the image borders is resolved computing 1D Haar wavelet in the vicinity of the borders. With the checkerboard splitting, at each level of decomposition only one detail image is produced that simplify the further analysis for data compression. The calculations are rather simple, without any floating point operation allowing the implementation of the designed algorithm in fixed point DSP processors for fast, near real time processing. The proposed algorithm does not possesses perfect restoration of the processed data because of rounding that is introduced at each level of decomposition/restoration to perform operations with byte represented data. The designed algorithm was tested on different images. The criterion to estimate quantitatively the quality of the restored images was the well known PSNR. For the visual quality estimation the error maps between original and restored images were calculated. The obtained simulation results show that the visual and quantitative quality of the restored images is degraded with number of decomposition level increasing but is sufficiently high even after 6 levels. The introduced distortion are concentrated in the vicinity of high spatial activity details and are absent in the homogeneous regions. The designed algorithm can be used for image lossy compression and in noise suppression applications.

In this paper we investigate the application of a rotational-shear interferometer, toward the problem of simultaneously estimating the directions of well-localized sources, and their spectral profiles. Rotational shear makes possible the acquisition of a spectrum estimate, without the mechanical scan required in using a Michelson interferometer in Fourier-transform spectroscopy. The spectrum and angle estimates are obtained computationally. The interferometric data enables the application of super-resolution direction-finding techniques commonly used in radar and sonar array processing.

An encrypted optical storage based on a joint transform correlator architecture by using a photorefractive material is presented. A key code designed by optimized algorithm so that its Fourier transform has a uniform amplitude distribution and a random phase distribution is introduced. Original two-dimensional data and the key code are placed side by side at the input plane. Both of them are stored in a photorefractive material as a joint power spectrum. The retrieval of the original data can be achieved with the same key code. We can record multiple two-dimensional data in a same crystal by angular multiplexing. Furthermore we can record by key code multiplexing because another designed key code can produce another independent phase distribution. We demonstrate the concept of the optical data storage system by computer simulations.

Correlation engines have been evolving since the implementation of radar. Here, correlation refers to association, which in this context is track-to-track association. In modern sensor fusion architectures, correlation and sensor registration are required to produce common, continuous, and unambiguous tracks of all objects in the surveillance area. The objective is to provide a unified picture of the theatre or area of interest to battlefield decision makers. This unified picture has many names, but is most commonly referred to as a Single Integrated Picture (SIP). A related process, known as sensor registration or gridlock filtering (gridlocking), refers to the reduction in navigation errors and sensor misalignment errors so that one sensor's track data can be accurately transformed into another sensor's coordinate system. As platforms gain multiple sensors, the correlation and gridlocking of tracks become significantly more difficult. Current correlation technology revolves around likelihood ratio theory and the assignment algorithm to resolve association ambiguities. While a Bayes classifier is the best classifier, all classifiers potentially lead to classification errors. In this paper, we examine the track association and sensor registration problem in terms of several correlation classifiers, the most famous of these being the matched filter. Thus, we seek some unification between the term correlation with regards to track association and correlation with regards to pattern recognition. We examine several classes of correlation classifiers and discuss their application to the generation of a SIP when coupled with a sensor registration algorithm. The availability of these techniques on optical processing platforms is an obvious benefit to track association. We briefly discuss the implementation of some of these techniques on a commercial frequency plane correlator.

The two-dimensional view, obtained with a camera, of a three-dimensional (3-D) object varies with the 3-D orientation of this object, complicating the recognition task. In this work we address the problem of estimating the pose of a 3-D object knowing only a 2-D projection. The proposed technique is based on a combination of synthetic-discriminant-function filters and neural networks. We succeed in estimating two orientations: in-plane and out-of-plane rotations within a 8 degree square range.