This PDF file contains the front matter associated with SPIE Proceedings Volume 8398, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

Detection and estimation of oil and oil-derived substances from an oil spill is a challenging issue. Over the last few years, several algorithms have been proposed for the detection of oil on the ocean surface. These techniques do not address the issue of detection of subsurface oil and estimate the depth of the location of oil at the subsurface level. In this paper, algorithms are developed to detect the presence of surface oil in ocean water using hyperspectral imagery. A support vector machine classifier was trained using region-of-interests (ROIs) to classify the oil/oil-derived substances under the water surface in the Gulf of Mexico. Using the pixel intensity of the identified oil based image, Beer-Lambert's law is used to calculate the depth at which the oil and/or oil-derived substance are present in the scene of investigation.

For humans, a picture is worth a thousand words, but to a machine, it is just a seemingly random array of numbers. Although machines are very fast and efficient, they are vastly inferior to humans for everyday information processing. Algorithms that mimic the way the human brain computes and learns may be the solution. In this paper we present a theoretical model based on the observation that images of similar visual perceptions reside in a complex manifold in an image space. The perceived features are often highly structured and hidden in a complex set of relationships or high-dimensional abstractions. To model the pattern manifold, we present a novel learning algorithm using a recurrent neural network. The brain memorizes information using a dynamical system made of interconnected neurons. Retrieval of information is accomplished in an associative sense. It starts from an arbitrary state that might be an encoded representation of a visual image and converges to another state that is stable. The stable state is what the brain remembers. In designing a recurrent neural network, it is usually of prime importance to guarantee the convergence in the dynamics of the network. We propose to modify this picture: if the brain remembers by converging to the state representing familiar patterns, it should also diverge from such states when presented with an unknown encoded representation of a visual image belonging to a different category. That is, the identification of an instability mode is an indication that a presented pattern is far away from any stored pattern and therefore cannot be associated with current memories. These properties can be used to circumvent the plasticity-stability dilemma by using the fluctuating mode as an indicator to create new states. We capture this behavior using a novel neural architecture and learning algorithm, in which the system performs self-organization utilizing a stability mode and an instability mode for the dynamical system. Based on this observation we developed a self- organizing line attractor, which is capable of generating new lines in the feature space to learn unrecognized patterns. Experiments performed on UMIST pose database and CMU face expression variant database for face recognition have shown that the proposed nonlinear line attractor is able to successfully identify the individuals and it provided better recognition rate when compared to the state of the art face recognition techniques. Experiments on FRGC version 2 database has also provided excellent recognition rate in images captured in complex lighting environments. Experiments performed on the Japanese female face expression database and Essex Grimace database using the self organizing line attractor have also shown successful expression invariant face recognition. These results show that the proposed model is able to create nonlinear manifolds in a multidimensional feature space to distinguish complex patterns.

In this paper, we describe a method for target detection using specialized masks for compressive sensing (CS) [1]. Unlike traditional CS, which uses random masks to sense a signal, we use basis functions that represent the targets of interest. Attention is given to the fact the location of the objects is not known, and the masks have to be shift invariant. Although this is similar to correlation filtering [2] in some respect, the intent is not to process a conventional image, but to directly project the scene on the mask to obtain measurements from which the information can be recovered. We illustrate this concept by describing a methodology for mask design that optimizes the detection performance, and presenting preliminary results of simulations.

The Support Vector Machine (SVM) is a powerful algorithm, useful in classifying data in to species. The SVMs implemented in this research were used as classifiers for the final stage in a Multistage Automatic Target Recognition (ATR) system. A single kernel SVM known as SVMlight, and a modified version known as a SVM with K-Means Clustering were used. These SVM algorithms were tested as classifiers under varying conditions. Image noise levels varied, and the orientation of the targets changed. The classifiers were then optimized to demonstrate their maximum potential as classifiers. Results demonstrate the reliability of SVM as a method for classification. From trial to trial, SVM produces consistent results.

Active imaging (AI) is necessary for measuring parameters of the objects that do not give out or reflect a specific type of radiation. AI systems offer a number of advantages over passive imaging systems that operate at visible through nearinfrared wavelengths and usually rely on solar illumination. The reliability and precision of the target identification depends on how the signal received from a sensor is processed. Often, obstacles or the imperfection of the sensors and processing electronics cause loss of some of the information. The technique of processes with missing data is suggested as part of time series prediction and analysis. Thus, the image may be reconstructed even if the necessary data is partially absent in the input signal. The suggested method reduces the false alarm rate of the target identification. Results are provided.

A computationally efficient appearance-based algorithm for geospatial object detection is presented and evaluated specifically for aircraft detection from satellite imagery. An aircraft operator exploiting the edge information via gray level differences between the aircraft and its background is constructed with Haar-like polygon regions by using the shape information of the aircraft as an invariant. Fast evaluation of the aircraft operator is achieved by means of integral image. Rotated integral images are utilized for detecting aircrafts in various orientations. Experiments are conducted on satellite images taken from various airport regions and promising results are obtained. Among tested various satellite images of 0.5 m resolution including 300 target aircrafts of various sizes, the proposed algorithm has resulted with typical values of 77% and 85% for precision and recall, respectively.

Interestingly, the past 20 years have provided us many examples of optical correlation methods for pattern recognition, e.g. VanderLugt correlator (VLC). In recent years, hybrid techniques, i.e. numerical implementation of correlation, have been also considered an alternative to all-optical methods because they show a good compromise between performance and simplicity. Moreover, these correlation methods can be implemented using an all-numerical and reprogrammable target such as the graphics processor unit (GPU), or the field-programmable gate array (FPGA). However, this numerical procedure requires realizing two Fourier Transforms (FT), a spectral multiplication, and a correlation plane analysis. The purpose of this study is to compare the performances of a numerical correlator based on the fast Fourier transform (FFT) with that relying on a simulation of the Fraunhofer diffraction. Different tests using the Pointing Head Pose Image Database (PHPID) and considering faces with vertical and horizontal rotations were performed with the code MATLAB. Tests were conducted with a five reference optimized composite filter. The receiving operating characteristics (ROC) curves show that the optical FT simulating the Fraunhofer diffraction leads to better performances than the FFT. The implications of our results for correlation are discussed.

Correlation is based pattern recognition is primarily based on the matching of contours between an unknown target image and a known reference image. However, it does not usually include the color image information in the decision making process. In order to render the correlation method sensitive to color change, we propose a generalized method based on the decomposition of the target image in its three color components using, either the normalized RGB (red, green, blue) color space, or the normalized HSV (hue, saturation, value) space. Then, the correlation operation is carried out for each color component and the results are merged in order to make a final decision. The aforementioned steps can alleviate majority of the problems associated with illumination changes in the target image by utilizing color information of the target image. To overcome these problems, we propose to convert the color based contour information into a signature corresponding to the color information of the target image. Test results are presented to validate the effectiveness of the proposed technique.

Optical joint transform correlation (JTC) has been proven to be an efficient pattern recognition tool, especially, for real-time applications. However, the classical JTC suffers from a lot of limitations such as broad correlation peaks, large side lobes, duplicate correlation peaks and low discrimination between target and non-target objects. This paper proposes a nonlinear JTC based target detection and tracking technique, where the reference image is phase-shifted and phase-encoded and then fed to two parallel processing channels. Each channel introduces the unknown input scene and performs Fourier transformations to obtain the joint power spectra signals, which are then combined and phase-encoded. Then a nonlinear operation is performed on the modified power spectrum followed by the application of fringe-adjusted filtering operation. A subsequent inverse Fourier transform operation yields the correlation output containing a highly distinct peak corresponding to each target present in the input scene. The reference image phase-encoding process removes any overlapping issue among the input scene objects, which is a drawback of classical JTC technique. An updated decision criterion is developed for the correlation plane so that it can accurately identify the location of the target. The proposed pattern recognition technique offers an excellent alternative for target tracking in an unknown video sequence.

In the article there are presented the mathematical and structural descriptions of the basic model of the optical correlator (BMOC), of the correlator using the matrixes of the lasers and filters (CMLF). In order to decrease the processing time in the correlators it is proposed to use the concept of the distribution of the operations of the targets detection and localization and to realize there in the different channels. At the stage of the targets detection it was proposed to use the filters generating the codified correlation functions consisting of a binary optical code which is analyzing in parallel with a high speed. There were elaborated new kinds of the correlators - with distributed targets detection and localization. There were given the analyses of the time expenditures and reliability in the different kinds of the correlators.

An improvement to the wavelet-modified Optimal Trade-off Maximum Average Correlation Height (OT-MACH) filter with the use of the Rayleigh distribution filter is proposed. The Rayleigh distribution filter is applied to the OT-MACH filter to provide a sharper low frequency cut-off than the Laplacian of Gaussian based wavelet filter that has been previously reported to enhance OT-MACH filter performance. Filters are trained using a 3D CAD model and tested on the corresponding real target object in high clutter environments acquired from a Forward Looking Infra Red (FLIR) sensor. Comparative evaluation of the performance of the original, wavelet and Rayleigh modified OT-MACH filter is reported for the recognition of the target objects present within the thermal infra-red image data set.

Automated Target Recognition (ATR) systems aim to automate target detection, recognition, and tracking. The current project applies a JPL ATR system to low-resolution sonar and camera videos taken from unmanned vehicles. These sonar images are inherently noisy and difficult to interpret, and pictures taken underwater were unreliable due to murkiness and inconsistent lighting. The ATR system breaks target recognition into three stages: 1) Videos of both sonar and camera footage are broken into frames and preprocessed to enhance images and detect Regions of Interest (ROIs). 2) Features are extracted from these ROIs in preparation for classification. 3) ROIs are classified as true or false positives using a standard Neural Network based on the extracted features. Several preprocessing, feature extraction, and training methods are tested and discussed in this report.

In prior work, we exploited the nonlinearity inherent in four-wave mixing in organic photorefractive materials for adaptive filtering. In this paper, we extend our work further and demonstrate new applications which involve: dislocation, scratches and defect enhancement. With the availability of the organic photorefractive materials with large space-bandwidth product, it should open the possibility of using the adaptive filtering techniques in quality control systems.

Application of distortion invariant filters (DIF) provides the possibility of invariant image recognition with increased speed of correlation matching. DIF with the minimization of correlation energy enable to control the properties of output correlation signal due to the parameterization during its synthesis. There are several types of such a filters presented nowadays. The relevance degree of each type of filter application is determined by the specific conditions of the recognition task. Thus it requires a comparative analysis of the filters performance. The simulations were provided for the DIF of the following types: MACE (Minimum Average Correlation Energy Filter), GMACE (Gaussian-minimum average correlation energy filters), MINACE (Minimum noise and correlation energy filter) and WMACE (the version of GMACE where the smoothing function is the wavelet). The synthesis of filters was carried out under identical conditions of gray-scale image recognition problem (out of plane rotated objects). The comparison results of discrimination characteristics and the requirements of DIFs synthesis are described and discussed.

Jet Propulsion Lab and Vescent Photonics Inc. and are jointly developing an innovative ultra-compact (volume < 10 cm³), ultra-low power (<10 ^-3 Watt-hours per measurement and zero power consumption when not measuring), completely non-mechanical Liquid Crystal Waveguide Fourier Transform Spectrometer (LCWFTS) that will be suitable for a variety of remote-platform, in-situ measurements. These devices are made possible by novel electro-evanescent waveguide architecture, enabling "monolithic chip-scale" Electro Optic-FTS (EO-FTS) sensors. The potential performance of these EO-FTS sensors include: i) a spectral range throughout 0.4-5 μm (25000 - 2000 cm-1), ii) highresolution (Δλ≤ 0.1 nm), iii) high-speed (< 1 ms) measurements, and iv) rugged integrated optical construction. This performance potential enables the detection and quantification of a large number of different atmospheric gases simultaneously in the same air mass and the rugged construction will enable deployment on previously inaccessible platforms. The sensor construction is also amenable for analyzing aqueous samples on remote floating or submerged platforms. We have reported [1] a proof-of-principle prototype LCWFTS sensor that has been demonstrated in the near- IR (range of 1450-1600 nm) with a 5 nm resolution. In this paper, we will report the recently built and tested LCWFTS test bed and the demonstration of a real-time gas sensing applications.

The objective of this paper is to develop a novel approach for encryption and compression of biometric information utilizing orthogonal coding and steganography techniques. Multiple biometric signatures are encrypted individually using orthogonal codes and then multiplexed together to form a single image, which is then embedded in a cover image using the proposed steganography technique. The proposed technique employs three least significant bits for this purpose and a secret key is developed to choose one from among these bits to be replaced by the corresponding bit of the biometric image. The proposed technique offers secure transmission of multiple biometric signatures in an identification document which will be protected from unauthorized steganalysis attempt.

Image sharpening is an image processing technique that highlights transitions in intensity and/or enhances the darker regions. This paper formulates a bidimensional empirical mode decomposition (BEMD) based spatial domain color image sharpening. In this approach, color image is first decomposed into several hierarchical components using BEMD, which is a multi-scale/multi-resolution technique. The hierarchical color image components are known as color bidimensional empirical mode functions (CBEMFs), where the first CBEMF contains the highest/finest local spatial variations, and the final CBEMF contains the color trend of an image. The final CBEMF is also known as color bidimensional residue (CBR), whereas the other CBEMFs are known as color bidimensional intrinsic mode functions (CBIMFs). However, instead of using classical BEMD, a modified BEMD, known as fast and adaptive BEMD (FABEMD) is utilized, which uses order-statics filters for envelope estimation in the process instead of surface interpolation. The BEMD developed for color images employing FABEMD is known as color BEMD (CBEMD). Since the first CBEMF contains the finest spatial variations in the image and the CBR contains the color trend information, manipulation of these two elements can provide useful sharpening of a color image. In one simple approach, suitable weighting of the first CBEMF and CBR is accomplished, where weighting is done to all three color components of these two elements. Finally, the image is reconstructed from the addition of all the CBEMFs to obtain the primary sharpening. An additional level of sharpening is achieved when the primarily sharpened image, as mentioned above, is added to the original image. By varying the weights, desired color image sharpening can be achieved, which is inherently data driven.

Digital steganography is gaining wide acceptance in the world of electronic copyright stamping. Digital media that are easy to steal, such as graphics, photos and audio files, are being tagged with both visible and invisible copyright stamps (known as digital watermarking). However, these same techniques can also be used to hide communications between actors in criminal or covert activities. An inherent difficulty in detecting steganography is overcoming the variety of methods for hiding a message and the multitude of choices of available media. Another problem in steganography defense is the issue of detection speed since the encoded data is frequently time-sensitive. When a message is visually transmitted in a non-textual format (i.e., in an image) it is referred to as a semagram. Semagrams are relatively easy to create, but very difficult to detect. While steganography can often be identified by detecting digital modifications to an image's structure, an image-based semagram is more difficult because the message is the image itself. The work presented describes the creation of a novel, computer-based application, which uses hybrid hierarchical neural network architecture to detect the likely presence of a semagram message in an image. The prototype system was used to detect semagrams containing Morse Code messages. Based on the results of these experiments our approach provides a significant advance in the detection of complex semagram patterns. Specific results of the experiments and the potential practical applications of the neural network-based technology are discussed. This presentation provides the final results of our research experiments.

In prior work, we demonstrate optical correlation via dynamic range compression in two-beam coupling using thin-film organic materials. In this paper, we continue the effort; characterize the performance of this correlator for variety of input. We successfully demonstrated correlation results almost free of cross- correlation and noise for extremely complicated noisy image were the signal image consist of several targets and reference image superposed of many templates.

Automatic speech processing systems are widely used in everyday life such as mobile communication, speech and speaker recognition, and for assisting the hearing impaired. In speech communication systems, the quality and intelligibility of speech is of utmost importance for ease and accuracy of information exchange. To obtain an intelligible speech signal and one that is more pleasant to listen, noise reduction is essential. In this paper a new Time Adaptive Discrete Bionic Wavelet Thresholding (TADBWT) scheme is proposed. The proposed technique uses Daubechies mother wavelet to achieve better enhancement of speech from additive non- stationary noises which occur in real life such as street noise and factory noise. Due to the integration of human auditory system model into the wavelet transform, bionic wavelet transform (BWT) has great potential for speech enhancement which may lead to a new path in speech processing. In the proposed technique, at first, discrete BWT is applied to noisy speech to derive TADBWT coefficients. Then the adaptive nature of the BWT is captured by introducing a time varying linear factor which updates the coefficients at each scale over time. This approach has shown better performance than the existing algorithms at lower input SNR due to modified soft level dependent thresholding on time adaptive coefficients. The objective and subjective test results confirmed the competency of the TADBWT technique. The effectiveness of the proposed technique is also evaluated for speaker recognition task under noisy environment. The recognition results show that the TADWT technique yields better performance when compared to alternate methods specifically at lower input SNR.

The modified matrix equivalently models (MMEMs) of multiport neural network heteroassociative memory (MP_NN_HAM) with double adaptive - equivalently weighing (DAEW) for recognition of 1D and 2D-patterns (images) are offered. It is shown, that computing process in MP_NN_HAM under using the proposed MMEMs, is reduced to two-step and multi-step algorithms and step-by-step matrix-matrix (tensor-tensor) procedures. The base operations and structural components for construction of MP_NN_HAM are matrix-matrix multipliers and matrixes of nonlinear converters, including threshold transformations. Advantages of such MMEMs for MP_NN_HAM were shown and confirmed by computer simulation results. The aim of paper is research of improved models and MP_NN_HAM for input 1D and 2D signals with unipolar coding and their capacity determination. The given results of computer simulations confirmed the perspective of such models. Results were also received for case of a MP_NN_HAM on base of MMEMs capacity exceeded a neurons amount. This memory is intended to recognize parallel and refresh P input distorted images (N-element vector). Such MP_NN_HAM is a kind of combination consisting of P independently functioning NN_HAM with common memory. Variants of optical realization of MP_NN_HAM architectures are considered in paper. A whole system is consists of two matrix-matrix (for 1D patterns) or two tensortensor (for 2D patterns) equivalentors (E) (or nonequivalentors (NE)) (MME and MMNE or TTE and TTNE).The proposed E (or NE) architecture with temporary integration has more large dimension of HAM and more simple design.

It is described a new image complexity informative feature, based on the input image Fourier spectrum calculation. There are presented the results of the experimental estimation of the proposed image complexity matrix. There are presented the results of the investigation of the influence of the image complexity on the required image resolution and of the influence of the image resolution and complexity on the correlation recognition.

The wavefront coding is a widely used in the optical systems to compensate aberrations and increase the depth of field. This paper presents experimental results on application of the wavefront coding paradigm for data encryption. We use a synthesised diffractive optical element (DOE) to deliberately introduce a phase distortion during the images registration process to encode the acquired image. In this case, an optical convolution of the input image with the point spread function (PSF) of the DOE is registered. The encryption is performed optically, and is therefore is fast and secure. Since the introduced distortion is the same across the image, the decryption is performed digitally using deconvolution methods. However, due to noise and finite accuracy of a photosensor, the reconstructed image is degraded but still readable. The experimental results, which are presented in this paper, indicate that the proposed hybrid optical-digital system can be implemented as a portable device using inexpensive off-the-shelf components. We present the results of optical encryption and digital restoration with quantitative estimations of the images quality. Details of hardware optical implementation of the hybrid optical-digital encryption system are discussed.

Local polar-edge detection (LPED) is a novel image processing method the author used in several computerized image systems in the last 3 years. It uses a novel edge-detection approach to detect the boundary points of a group of temperature-selected objects embedded in a large IR image frame. Then it uses a 2-D clustering method to group all boundary points into N sub-groups with each sub-group representing one particular high-temperature object. It will then be followed by finding the center of mass point, or CMP, of each sub-group. From each sub- group, the program will immediately find 36 radial distances between the CMP and the boundary. A 36-dimension analog vector can then be constructed from these 36 radial distances. This 36D analog vector is the ID vector to identify the object that has this particular boundary. This ID vector is independent of the object location and independent of the object orientation. But it is a unique property from object to object. Therefore it can be used to track and target any particularly shaped object, especially when the object is moving and the focusing is not very sharp. All these image-processing steps are automatically carried out in a super-fast speed, once when the IR- image is loaded into the program. Because of the super-short time used to implement the novel algorithm, automatic tracking and automatic targeting can therefore be carried out in real-time.

Wavefront coding paradigm can be used not only for compensation of aberrations and depth-of-field improvement but also for an optical encryption. An optical convolution of the image with the PSF occurs when a diffractive optical element (DOE) with a known point spread function (PSF) is placed in the optical path. In this case, an optically encoded image is registered instead of the true image. Decoding of the registered image can be performed using standard digital deconvolution methods. In such class of optical-digital systems, the PSF of the DOE is used as an encryption key. Therefore, a reliability and cryptographic resistance of such an encryption method depends on the size and complexity of the PSF used for optical encoding. This paper gives a preliminary analysis on reliability and possible vulnerabilities of such an encryption method. Experimental results on brute-force attack on the optically encrypted images are presented. Reliability estimation of optical coding based on wavefront coding paradigm is evaluated. An analysis of possible vulnerabilities is provided.