This PDF file contains the front matter associated with SPIE Proceedings Volume 8748, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

We look at the design of projective measurements for compressive imaging based upon image priors and device constraints. If one assumes that image patches from natural imagery can be modeled as a low rank manifold, we develop an optimality criterion for a measurement matrix based upon separating the canonical elements of the manifold prior. We then characterize this manifold based upon prior training imagery under a treebased framework which can be implemented adaptively. We also illustrate how these adaptive measurements can incorporate prior knowledge regarding the constrains of the device being used to collect the measurements. Simulated performance results are presented and compared against a standard imaging paradigm as well as more conventional compressive imaging techniques.

Jet Propulsion Laboratory has developed an innovative Grayscale Optical Correlator (GOC) architecture using a pair of Digital Light Processor Spatial Light Modulator (DLP SLM) as the input and filter devices and a CMOS sensor for correlation output detection [1-5]. In order to achieve ultra high-speed Automatic Target Recognition (ATR), we have developed custom Electronic Interfaces to maximize the system data throughput rate for both the DLP and CMOS. The high-performance Electronic Interface System (EIS) is capable of achieving sustained 1000 frames per second (fps) at 1920x1024 data frame size. In this paper, we will first overview the new GOC architecture. We will the depict the detailed design of the EIS for the DLP SLM and CMOS. The innovation of JPL’s high-performance digital/optical ATR system is in its implementation of a high-speed, high-resolution DLP display, and a high-speed CMOS camera sensor with an advanced I/O interface and high-speed parallel on-board processing capability.

The science fiction inspired shrinking of macro-scale robotic manipulation and handling down to the micro- and nanoscale regime opens new doors for exploiting the forces and torques of light for micro- and nanoscopic probing, actuation and control. Advancing light-driven micro-robotics requires the optimization of optical forces and torques that, in turn, requires optimization of the underlying light-matter interaction. This report is two-fold desribing the new use of proprietary strongholds we currently are harnessing in the Programmable Phase Optics in Denmark on new means of sculpting of both light and matter for robotically probing at the smallest biological length scales.

Optical structures to implement the discrete Fourier transform (DFT) and fast Fourier transform (FFT) algorithms for discretely sampled data sets are considered. In particular, the decomposition of the FFT algorithm into the basic Butterfly operations is described, as this allows the algorithm to be fully implemented by the successive coherent addition and subtraction of two wavefronts (the subtraction being performed after one has been appropriately phase shifted), so facilitating a simple and robust hardware implementation based on waveguided hybrid devices as employed in coherent optical detection modules. Further, a comparison is made to the optical structures proposed for the optical implementation of the quantum Fourier transform and they are shown to be very similar.

In this study, we suggest and validate an all-numerical implementation of a VanderLugt correlator which is optimized for face recognition applications. The main goal of this implementation is to take advantage of the benefits (detection, localization, and identification of a target object within a scene) of correlation methods and exploit the reconfigurability of numerical approaches. This technique requires a numerical implementation of the optical Fourier transform. We pay special attention to adapt the correlation filter to this numerical implementation. One main goal of this work is to reduce the size of the filter in order to decrease the memory space required for real time applications. To fulfil this requirement, we code the reference images with 8 bits and study the effect of this coding on the performances of several composite filters (phase-only filter, binary phase-only filter). The saturation effect has for effect to decrease the performances of the correlator for making a decision when filters contain up to nine references. Further, an optimization is proposed based for an optimized segmented composite filter. Based on this approach, we present tests with different faces demonstrating that the above mentioned saturation effect is significantly reduced while minimizing the size of the learning data base.

3D image reconstruction is desirable in many applications such as city planning, cartography and many vision applications. The accuracy of the 3D reconstruction plays a vital role in many real world applications. We introduce a method which uses one LiDAR image and N conventional visual images to reduce the error and to build a robust registration for 3D reconstruction. In this method we used lines as features in both the LiDAR and visual images. Our proposed system consists of two steps. In the first step, we extract lines from the LiDAR and visual images using Hough transform. In the second step, we estimate the camera matrices using a search algorithm combined with the fundamental matrices for the visual cameras. We demonstrate our method on a synthetic model which is an idealized representation of an urban environment.

The purpose of this paper is to test correlation methods for pattern recognition applications. A broad overview of the main correlation architectures is first given. Many correlation data are compared with those obtained from standard pattern recognition methods. We used our simulations to predict improved decisional performance from correlation methods. More specifically, we are focused on the POF filter and composite filter family. We present an optimized composite correlation filter, called asymmetric segmented phase-only filter (ASPOF) for mobile target recognition applications. The main objective is to find a compromise between the number of references to be merged in the correlation filter and the time needed for making a decision. We suggest an all-numerical implementation of a VanderLugt (VLC) type composite filter. The aim of this all-numerical implementation is to take advantage of the benefits of the correlation methods and make the correlator easily reconfigurable for various scenarios. The use of numerical implementation of the optical Fourier transform improves the decisional performance of the correlator. Further, it renders the correlator less sensitive to the saturation phenomenon caused by the increased number of references used for fabricating the composite filter. Different tests are presented making use of the peak-to-correlation energy criterion and ROC curves. These tests confirm the validity ofour technique. Elderly fall detection and underwater mine detection are two applications which are considered for illustrating the benefits of our approach. The present work is motivated by the need for detailed discussions of the choice of the correlation architecture for these specific applications, pre-processing in the input plane and post processing in the output plane techniques for such analysis.

An extension of the recently proposed method of simultaneous compression and encryption of multiple images [Opt. Lett. 35, 1914-1916 (2010)] is developed. This analysis allows us to find a compromise between compression rate and quality of the reconstructed images for target detection applications. This spectral compression method can significantly reduce memory size and can be easily implemented with a VanderLugt correlator (VLC). For that purpose, we determine the size of the useful spectra for each target image by exploiting the root-mean-square time-frequency criterion. This parameter is used to determine the allowed area of each target image within the compressed spectrum. Moreover, this parameter is adapted in order to minimize overlapping between the different spectra. For that purpose we add a shift function adapted to each spectra. Finally, the spectra are merged together by making use of a segmentation criterion. The latter compares the local energy relative to each pixel for each spectrum. Furthermore, it optimizes assignment of the considered pixel by taking into account the adjacent areas to the considered pixel. This permits to avoid the presence of isolated areas and small sized areas (less than 10 pixels). In this paper, we analyse and optimize the shift function needed to separate the different spectra. We use mean square error (MSE) for comparing compression rates. A series of tests with several video sequences show the benefit of this shift function on the quality of reconstructed images and compression rate.

Digital elevation models (DEMs) are a digital representation of elevations at regularly spaced points. They provide an accurate tool to extract the shoreline profiles. One of the emerging sources of creating them is light detection and ranging (LiDAR) that can capture a highly dense cloud points with high resolution that can reach 15 cm and 100 cm in the vertical and horizontal directions respectively in short periods of time. In this paper we present a multi-step morphological algorithm to extract shorelines locations from the DEM data and a predefined tidal datum. Unlike similar approaches, it utilizes Lowess nonparametric regression to estimate the missing values within the DEM file. Also, it will detect and eliminate the outliers and errors that result from waves, ships, etc by means of anomality test with neighborhood constrains. Because, there might be some significant broken regions such as branches and islands, it utilizes a constrained morphological open and close to reduce these artifacts that can affect the extracted shorelines. In addition, it eliminates docks, bridges and fishing piers along the extracted shorelines by means of Hough transform. Based on a specific tidal datum, the algorithm will segment the DEM data into water and land objects. Without sacrificing the accuracy and the spatial details of the extracted boundaries, the algorithm should smooth and extract the shoreline profiles by tracing the boundary pixels between the land and the water segments. For given tidal values, we qualitatively assess the visual quality of the extracted shorelines by superimposing them on the available aerial photographs.

Image registration is basic step in image fusion and in many 2D applications. Registering the 2D image with recent robust algorithms like SIFT (Scale invariant Feature Transform) works well in most situations. However, registering the 2D images under poor illumination is a challenging problem. In several situations, conventional registration algorithms like SIFT fail to register the images. Aside from poor illumination conditions, images involving too much symmetry can also pose registration difficulties for conventional methods. In our approach, we overcome these limitations by using the knowledge of the intrinsic camera parameters together with a new registration method to help in registering the features (lines) between the two overlapping images. Our approach is useful especially in registering the images taken by different sensors or the same sensor at different times under poor illumination conditions. Experiments are tested on real world environments.

The speech signal of a word is a combination of frequencies which can produce specific transition frequency shapes. These can be regarded as a written text in some unknown ‘script’. Before attempting methods to read the speech spectrogram image using image processing techniques we need first to define the properties of the speech spectrogram image as well as the reduction of the clutter of the spectrogram image and the selection of the methods to be employed for image matching. Thus methods to convert the speech signal to a spectrogram image are initially employed, followed by reduction of the noise in the signal by capturing the energy associated with formants of the speech signal. This is followed by the normalisation of the size of the image and its resolution of in both the frequency and time axes. Finally, template matching methods are employed to recognise portions of text and isolated words. The paper describes the pre-processing methods employed and outlines the use of normalised grey-level correlation for the recognition of words.

An image hiding method based on cascaded iterative Fourier transform and public-key encryption algorithm was proposed. Firstly, the original secret image was encrypted into two phase-only masks M1 and M2 via cascaded iterative Fourier transform (CIFT) algorithm. Then, the public-key encryption algorithm RSA was adopted to encrypt M2 into M2' . Finally, a host image was enlarged by extending one pixel into 2×2 pixels and each element in M1 and M2' was multiplied with a superimposition coefficient and added to or subtracted from two different elements in the 2×2 pixels of the enlarged host image. To recover the secret image from the stego-image, the two masks were extracted from the stego-image without the original host image. By applying public-key encryption algorithm, the key distribution was facilitated, and also compared with the image hiding method based on optical interference, the proposed method may reach higher robustness by employing the characteristics of the CIFT algorithm. Computer simulations show that this method has good robustness against image processing.

In this paper, an iterative tracking algorithm based on a non-linear JTC (Joint Transform Correlator) architecture and enhanced by a digital image processing method is proposed and validated. This algorithm is based on the computation of a correlation plane where the reference image is updated at each frame. For that purpose, we use the JTC technique in real time to track a patient (target image) in a room fitted with a video camera. The correlation plane is used to localize the target image in the current video frame (frame i). Then, the reference image to be exploited in the next frame (frame i+1) is updated according to the previous one (frame i). In an effort to validate our algorithm, our work is divided into two parts: (i) a large study based on different sequences with several situations and different JTC parameters is achieved in order to quantify their effects on the tracking performances (decimation, non-linearity coefficient, size of the correlation plane, size of the region of interest...). (ii) the tracking algorithm is integrated into an application of elderly fall detection. The first reference image is a face detected by means of Haar descriptors, and then localized into the new video image thanks to our tracking method. In order to avoid a bad update of the reference frame, a method based on a comparison of image intensity histograms is proposed and integrated in our algorithm. This step ensures a robust tracking of the reference frame. This article focuses on face tracking step optimisation and evalutation. A supplementary step of fall detection, based on vertical acceleration and position, will be added and studied in further work.

As a uncommon biometric modality, human gait recognition has a great advantage of identify people at a distance without high resolution images. It has attracted much attention in recent years, especially in the fields of computer vision and remote sensing. In this paper, we propose a human gait recognition framework that consists of a reliable background subtraction method followed by the pyramid of Histogram of Gradient (pHOG) feature extraction on the silhouette image, and a Hidden Markov Model (HMM) based classifier. Through background subtraction, the silhouette of human gait in each frame is extracted and normalized from the raw video sequence. After removing the shadow and noise in each region of interest (ROI), pHOG feature is computed on the silhouettes images. Then the pHOG features of each gait class will be used to train a corresponding HMM. In the test stage, pHOG feature will be extracted from each test sequence and used to calculate the posterior probability toward each trained HMM model. Experimental results on the CASIA Gait Dataset B1 demonstrate that with our proposed method can achieve very competitive recognition rate.

A novel and robust technique is proposed in this paper for securing confidential information by utilizing orthogonal coding scheme, encoded steganography and nonlinear encryption through joint transform correlation. Different biometric signatures are encoded using individual orthogonal codes and then multiplexed together. The encrypted and multiplexed image is hidden inside a cover image employing a steganography technique, where one from the three least significant bits is chosen using another secret key. A color cover image is utilized which is decomposed into three color components, red, green and blue, so that three different sets of biometric signatures can be embedded into each of the color components. The color stego image is finally encrypted using multiple phase-shifted reference joint transform correlation (MRJTC) technique. The proposed encryption technique is a nonlinear process which increases the security strength significantly against any unauthorized access. The encoded steganography technique reduces the vulnerability that an intruder can retrieve any information from a given image through any steganalysis attack. Finally, the orthogonal coding scheme enhances the robustness by making the biometric information almost inaccessible without authorization.

Automatic pattern recognition algorithms are implemented to correct distortion and remove noise from the optical medium in the multi-channel optical communication systems. The post-processing involves filtering and correlation to search for accurate location of every optical data element. Localized thresholding and neural network training methods are used to accurately digitize the analog optical images into digital data pages. The goal is to minimize the bit-errorrate (BER) in the optical data transmission and receiving process. Theoretical analysis and experimental tests have been carried out to demonstrate the improved optical data retrieval accuracy.

This paper presents an approach related to automated recognition of small features of movable targets including fast moving objects such as airplanes, etc. Small features recognition is a challenging problem in both fields: pattern recognition of particular configurations and of complexes comprising a number of configurations. Specific target details, although well characterized by their features are often arranged in an elaborated way which makes the recognition task very difficult and welcomes new ideas (approaches). On the other hand, the variety of small characters (features) is intrinsically linked to the technology development of the identified targets and is unavoidable. Due to the complexity of possible technological designs, the feature representation is one of the key issues in optical pattern recognition. A flexible hierarchical prediction modeling is proposed with application examples.

The paper presents the results of computer modeling of scaled images recognition using MACE, GMACE and MINACE invariant correlation filters. There is given data about testing of above mentioned filters on database of grayscale images with different resolution that contains only true class and both true class and false class objects. There presented data about output correlation peak qualitative and quantitative characteristics and about the comparison for different filter types. The filters were synthesized specially for case of such geometrical transform as change of scale. Also there is presented data about testing of mentioned filters for recognition of rotated images and given an analysis of results.

Optical correlators are well known to be perspective for real time image recognition. Application of distortion invariant filters (DIF) provides image recognition with increased speed of correlation image matching. Minimum noise and correlation energy filters (MINACE filters) provide good recognition in the case of gray-scale input images. These filters possess a good mathematical basis and can be efficiently implemented in digital processing systems or in hybrid opto-digital correlators at a high rate. This paper is subjected to synthesis and realization of MINACE filters for 4-f correlator as computer generated holograms (holographic filters).

Security holograms (SH) are perspective for document and product authenticity protection due to difficulties of such a protection mark falsification. Mass production of SH uses widespread technology of hot foil or lavsan paper stamping. The quality of holograms significantly depends on perfection of nickel master-matrix that is used in stamping equipment. We represent the method of automatic quality inspection of nickel master-matrix based on digital processing of its surface relief microphotographs. Proposed processing algorithm is based on combination of image spatial frequency analysis and image matching using distortion invariant correlation filters. The results of our method application for real SH master-matrices inspection are shown in this paper.

An efficient unsupervised technique is proposed for land mine detection from highly cluttered inhomogeneous environment. The proposed technique uses multispectral data for which feature extraction is necessary to classify large volume of data. We applied wavelet based principal component analysis to reduce the dimension of the data as well as to reveal information about target from background clutter. To increase the discrimination between target and clutter a linear transformation of the feature extracted bands is performed. Thereafter, morphological algorithm is used to extract the maximum information about the target. The proposed technique shows excellent detection performance while enhancing the processing speed. Test results using various multispectral data sets show excellent performance and verify the effectiveness of the proposed technique.

Detection of concealed objects under cloth or inside paper/lather/plastic box is a challenge for security applications. With terahertz (THz) imaging technology, it is possible to spot concealed objects inside plastic box, underneath cloths paper or similar scenarios. THz frequency domain (~100 GHz - 10 THz) shows a unique feature in the under-used domain of the electromagnetic spectrum which helps to acquire image of concealed objects. This property of THz wave makes it useful in a variety of applications. Previously millimeter wave imaging and infrared imaging were used for detection of concealed features in an image with limited success rate. THz imaging helps solving the problem to a great extent because it can transmit through substances like cloths, paper, plastic, dried food etc. THz images have poor quality and low signal-to-noise-ratio. Noises and related artifacts must be reduced for proper detection of concealed objects. In this paper, a new technique for artifact reduction and detection of concealed object is proposed by utilizing nonzero-order fringe adjusted joint transform correlation (NFJTC) technique. In the proposed NFJTC technique, the joint power spectrum (JPS) is modified to obtain the nonzero-order fringe-adjusted joint power spectrum. NFJTC is already been used for object detection but never been used to detect concealed objects in THz imagery. Test results using real life THz imagery confirm the effectiveness of the proposed technique.

Dim small target detection, which is characterized by complex background and low Signal-to-Noise Ratio (SNR), is critical for many applications. Traditional detection algorithms assume that noise is not useful for detecting targets and try to remove the noise to improve SNR of images using various filtering techniques. In this paper, we introduce a detection algorithm based on Stochastic Resonance (SR) where stochastic resonance is used to enhance the dim small targets. Our intuition is that SR can achieve the target enhancement in the presence of noise. Adaptive Least Mean Square (ALMS) filtering is first adopted to estimate the background, and the clutter is suppressed by subtracting the estimated background image from the source image. Adaptive SR (ASR) method is then employed to enhance the target and improve the SNR of the image containing the target and noise. ASR tunes and adds the optimal noise intensity to increase the power of the targets and therefore improve the SNR of the image. Several experiments on synthetic and natural images are conducted to evaluate our proposed algorithm. The results demonstrate the effectiveness of our algorithm.

The spectral fringe-adjusted joint transform correlation (SFJTC) has been used effectively for performing deterministic target detection in hyperspectral imagery. However, experiments show decreased performance when noise-corrupted spectral variability is present in the target signatures. In this paper, we propose to use a modified spectral fringe-adjusted joint transform correlation based target detection algorithm, which employs a new real-valued filter called the logarithmic fringe-adjusted filter (LFAF). Furthermore, the maximum noise fraction (MNF) technique is used for preprocessing the hyperspectral imagery, which makes the SFJTC technique more insensitive to spectral variability in noisy environment. Test results using real life oil spill based hyperspectral image datasets show that the proposed scheme yields better performance compared to alternate techniques.