This PDF file contains the front matter associated with SPIE Proceedings Volume 6978, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and the Conference Committee listing.

A Gaussian model-based statistical matching procedure is proposed for image enhancement and segmentation. Generally speaking, enhanced images are desired for visual analysis whereas segmented images are required for target recognition. A histogram matching procedure is used to enhance a given image. To perform histogram matching, two histograms are needed, an original histogram computed from the given image and a specified histogram to be matched to. For image enhancement, the specified histogram is a Gaussian model (mean & standard deviation) that can be estimated from a number of well-exposed images or properly processed images. Certainly the Gaussian model varies with the category of imagery. For image segmentation, N Gaussian models (means & standard deviations) are estimated from the original histogram of a given image. The number of Gaussian models (N) is decided by analyzing the original histogram. A statistical matching procedure is used to map the original histogram onto one of the Gaussian models defined by their means and standard deviations. Specifically, the mapped image can be computed by subtracting the mean of original image from the original image, scaling with the ratio of the standard deviation of Gaussian model to the standard deviation of original image and plus the mean of Gaussian model. The statistically mapped image is thresheld by using the mean of Gaussian model, which results one set of expected segments. The statistical matching plus thresholding procedure is repeated N times for N Gaussian models. Finally, all N sets of segments are fully obtained. The proposed image enhancement and segmentation procedure are validated with multi-sensor imagery.

An adaptive technique for image enhancement based on a specifically designed nonlinear function is presented in this paper. The enhancement technique constitutes three main processes-adaptive intensity enhancement, contrast adjustment, and color restoration. A sine function with an image dependent parameter is used to tune the intensity of each pixel in the luminance image. This process provides dynamic range compression by boosting the luminance of darker pixels while reducing the intensity of brighter pixels and maintaining local contrast. The normalized reflectance image is added to the enhanced image to preserve the details. The quality of the enhanced image is improved by applying a local contrast enhancement followed by a contrast stretch process. A basic linear color restoration process based on the chromatic information of the original image is employed to convert the enhanced intensity image back to a color image. The performance of the algorithm is compared with other state of the art enhancement techniques and evaluated using a statistical image quality evaluation method.

It is critical in military applications to be able to extract features in imagery that may be of interest to the viewer at any time of the day or night. Infrared (IR) imagery is ideally suited for producing these types of images. However, even under the best of circumstances, the traditional approach of applying a global automatic gain control (AGC) to the digital image may not provide the user with local area details that may be of interest. Processing the imagery locally can enhance additional features and characteristics in the image which provide the viewer with an improved understanding of the scene being observed. This paper describes a multi-resolution pyramid approach for decomposing an image, enhancing its contrast by remapping the histograms to desired pdfs, filtering them and recombining them to create an output image with much more visible detail than the input image. The technique improves the local area image contrast in light and dark areas providing the warfighter with significantly improved situational awareness.

In this paper, a new wavelet-based dynamic range compression algorithm is proposed to improve the visual quality of digital images captured in the high dynamic range scenes with non-uniform lighting conditions. Wavelet transform is used especially for dimension reduction such that a dynamic range compression with local contrast enhancement algorithm is applied only to the approximation coefficients which are obtained by low-pass filtering and down-sampling the original intensity image. The normalized approximation coefficients are transformed using a hyperbolic sine curve and the contrast enhancement is realized by tuning the magnitude of the each coefficient with respect to surrounding coefficients. The transformed coefficients are then de-normalized to their original range. The detail coefficients are also modified to prevent the edge deformation. The inverse wavelet transform is carried out resulting in a low dynamic range and contrast enhanced intensity image. A color restoration process based on relationship between spectral bands and the luminance of the original image is applied to convert the enhanced intensity image back to a color image.

A modular approach on an adaptive thresholding method for segmentation of cell regions in bioelectric images with complex lighting environments and background conditions is presented in this paper. Preprocessing steps involve lowpass filtering of the image and local contrast enhancement. This image is then adaptively thresholded which produces a binary image. The binary image consists of cell regions and the edges of a metal electrode that show up as bright spots. A local region based approach is used to distinguish between cell regions and the metal electrode tip that cause bright spots. Regional properties such as area are used to separate the cell regions from the non-cell regions. Special emphasis is given on the detection of twins and triplet cells with the help of watershed transformation, which might have been lost if form-factor alone were to be used as the geometrical descriptor to separate the cell and the non-cell regions.

The performance measurement of hypervelocity projectiles in flight is critical in ensuring proper projectile operation, for designing new long-range missile systems with improved accuracy, and for assessing damage to the target upon impact to determine the projectile's lethality. We are developing a modular, low cost, digital X-ray imaging system to measure hypervelocity projectile parameters with high precision and to almost instantaneously map its trajectory in 3D space to compute its pitch, yaw, displacement from its path, and velocity. The preliminary data suggest that this system can render an accuracy of 0.25° in measuring pitch and yaw, an accuracy of 0.03" in estimating displacement from the centerline, and a precision of ±0.0001% in measuring velocity, which is well beyond the capability of any existing system.

In this paper, we present a region-of-interest-based video coding system for use in real-time applications. Region-of-interest (ROI) coding methodology specifies that targets or ROIs be coded at higher fidelity using a greater number of available bits, while the remainder of the scene or background is coded using a fewer number of bits. This allows the target regions within the scene to be well preserved, while dramatically reducing the number of bits required to code the video sequence, thus reducing the transmission bandwidth and storage requirements. In the proposed system, the ROI contours can be selected arbitrarily by the user via a graphical user interface (GUI), or they can be specified via a text file interface by an automated process such as a detection/tracking algorithm. Additionally, these contours can be specified at either the transmitter or receiver. Contour information is efficiently exchanged between the transmitter and receiver and can be adjusted on the fly and in real time. Coding results are presented for both electro-optical (EO) and infrared (IR) video sequences to demonstrate the performance of the proposed system.

This paper describes a registration algorithm for aligning large frame imagery compressed with the JPEG2000 compression standard. The images are registered in the compressed domain using wavelet-based techniques. Unlike traditional approaches, our proposed method eliminates the need to reconstruct the full image prior to performing registration. The proposed method is highly scalable allowing registration to be performed on selectable resolution levels, quality layers, and regions of interest. The use of the hierarchical nature of the wavelet transform also allows for the trade-off between registration accuracy and processing speed. We present the results from our simulations to demonstrate the feasibility of the proposed technique in real-world scenarios with streaming sources. The wavelet-based approach maintains compatibility with JPEG2000 and enables additional features not offered by traditional approaches.

Comparing two similar images is often needed to evaluate the effectiveness of an image processing algorithm. But, there is no one widely used objective measure. In many papers, the mean squared error (MSE) or peak signal to noise ratio (PSNR) are used. These measures rely entirely on pixel intensities. Though these measures are well understood and easy to implement, they do not correlated well with perceived image quality. This paper will present an image quality metric that analyzes image structure rather than entirely on pixels. It extracts image structure with the use of a recursive quadtree decomposition. A similarity comparison function based on contrast, luminance, and structure will be presented.

The general image quality equation (GIQE) [Leachtenauer et al., Appl. Opt. 36, 8322-8328 (1997)] is an empirical formula for predicting the quality of imagery from a given incoherent optical system in terms of the National Imagery Interpretability Rating Scale (NIIRS). In some scenarios, the two versions of the GIQE (versions 3.0 and 4.0) yield significantly different NIIRS predictions. We compare visual image quality assessments for simulated imagery with GIQE predictions and analyze the physical basis for the GIQE terms in an effort to determine the proper coefficients for use with Wiener-filtered reconstructions of Nyquist and oversampled imagery in the absence of aberrations. Results indicate that GIQE 3.0 image quality predictions are more accurate than those from GIQE 4.0 in this scenario.

We will review recent developments in coded aperture techniques for unconventional imaging systems. Specifically, we are interested in looking simultaneously in multiple directions using a common aperture. To accomplish this, we interleave several sparse sub-apertures that are pointed in different directions. The goal is to optimize the sub-apertures so that the point spread function (PSF) is well behaved, and resolution is preserved in the images. We will present an analysis of the underlying PSF design concept, as well as the necessary phase optimization techniques.

Computational imaging systems are characterized by a joint design and optimization of front end optics, focal plane arrays and post-detection processing. Each constituent technology is characterized by its unique scaling laws. In this paper we will attempt a synthesis of the behavior of individual components and develop scaling analysis of the jointly designed and optimized imaging systems.

Spectral imaging is an emerging tool for defense and security applications because it provides compositional information about the objects in a scene. The underlying task-measuring a 3-D dataset using a 2-D detector array-is challenging, and straightforward approaches to the problem can result in severe performance tradeoffs. While a number of ingenious (non-adaptive) solutions have been proposed that minimize these tradeoffs, the complexity of the sensing task suggests that adaptive approaches to spectral imaging are worth considering. As a first step towards this goal, we investigate adaptive spectroscopy and present initial results confirming dramatic cost/performance gains for a particular implementation.

Compressive Sensing (CS) is a recently emerged signal processing method. It shows that when a signal is sparse in a certain basis, it can be recovered from a small number of random measurements made on it. In this work, we investigate the possibility of utilizing CS to sample the video stream acquired by a fixed surveillance camera in order to reduce the amount of data transmitted. For every 15 continuous video frames, we select the first frame in the video stream as the reference frame. Then for each following frame, we compute the difference between this frame and its preceding frame, resulting in a difference frame, which can be represented by a small number of measurement samples. By only transmitting these samples, we greatly reduce the amount of transmitted data. The original video stream can still be effectively recovered. In our simulations, SPGL1 method is used to recover the original frame. Two different methods, random measurement and 2D Fourier transform, are used to make the measurements. In our simulations, the Peak Signal-to-Noise Ratio (PSNR) ranges from 28.0dB to 50.9dB, depending on the measurement method and number of measurement used, indicating good recovery quality. Besides a good compression rate, the CS technique has the properties of being robust to noise and easily encrypted which all make CS technique a good candidate for signal processing in communication.

The well known phase diversity technique has long been used as a premier passive imaging method to mitigate the degrading effects of atmospheric turbulence on incoherent optical imagery. Typically, an iterative, slow method is applied that uses the Zernike basis set and 2-D Fourier transforms in the reconstruction process. In this paper, we demonstrate a direct method for estimating the un-aberrated object brightness from phase or phase difference estimates that 1) does not require the use of the Zernike basis set or the intermediate determination of the generalized pupil function, 2) directly determines the optical transfer function without the requirement for an iterative sequence of 2-D Fourier Transforms, 3) provides a more accurate result than the Zernike-based approaches since there are no Zernike series truncation errors, 4) lends itself to fast and parallel implementation, and 5) can use stochastic search methods to rapidly determine simultaneous phases or phase differences required to determine the correct optical transfer function estimate. As such, this new implementation of phase diversity provides potentially faster, more accurate results than previous approaches yet still retains inherent compatibility with the traditional Zernike-based methods. The theoretical underpinnings of this new method along with demonstrative computer simulation results are presented.

A fundamental element of future generic pattern recognition technology is the ability to extract similar patterns for the same scene despite wide ranging extraneous variables, including lighting, turbidity, sensor exposure variations, and signal noise. In the process of demonstrating pattern constancy of this kind for retinex/visual servo (RVS) image enhancement processing, we found that the pattern constancy performance depended somewhat on scene content. Most notably, the scene topography and, in particular, the scale and extent of the topography in an image, affects the pattern constancy the most. This paper will explore these effects in more depth and present experimental data from several time series tests. These results further quantify the impact of topography on pattern constancy. Despite this residual inconstancy, the results of overall pattern constancy testing support the idea that RVS image processing can be a universal front-end for generic visual pattern recognition. While the effects on pattern constancy were significant, the RVS processing still does achieve a high degree of pattern constancy over a wide spectrum of scene content diversity, and wide ranging extraneousness variations in lighting, turbidity, and sensor exposure.

The purpose of this research is to examine how to achieve suitable aggregation in the simulation of large systems. More specifically, investigating how to accurately aggregate hierarchical lower-level (higher resolution) models into the next higher-level in order to reduce the complexity of the overall simulation model. The initial approach used in this research was to use a realistic simulation model of a complex flying training model to apply the model aggregation methodologies using artificial neural networks and other statistical techniques. In order to test the techniques proposed, we modified a flying training model built for another study to suit the needs of our experiment. The study examines the effectiveness of three types of artificial neural networks as a metamodel in predicting outputs of the flying training model. Feed-forward, radial basis function, and generalized regression neural networks are considered and are compared to the truth simulation model, where the truth model is when actual lower-level model outputs are used as a direct input into the next higher-level model. The desired real-world application of the developed simulation aggregation process will be applied to military combat modeling in the area of combat identification (CID).

Adaptive methods are defined and experimentally studied for a two-scale edge detection process that mimics human visual perception of edges and is inspired by the parvo-cellular (P) and magno-cellular (M) physiological subsystems of natural vision. This two-channel processing consists of a high spatial acuity/coarse contrast channel (P) and a coarse acuity/fine contrast (M) channel. We perform edge detection after a very strong non-linear image enhancement that uses smart Retinex image processing. Two conditions that arise from this enhancement demand adaptiveness in edge detection. These conditions are the presence of random noise further exacerbated by the enhancement process, and the equally random occurrence of dense textural visual information. We examine how to best deal with both phenomena with an automatic adaptive computation that treats both high noise and dense textures as too much information, and gracefully shifts from a smallscale to medium-scale edge pattern priorities. This shift is accomplished by using different edge-enhancement schemes that correspond with the (P) and (M) channels of the human visual system. We also examine the case of adapting to a third image condition, namely too little visual information, and automatically adjust edge detection sensitivities when sparse feature information is encountered. When this methodology is applied to a sequence of images of the same scene but with varying exposures and lighting conditions, this edge-detection process produces pattern constancy that is very useful for several imaging applications that rely on image classification in variable imaging conditions.

We describe an approach to automatically detect building facades in images of urban environments. This is an important problem in vision-based navigation, landmark recognition, and surveillance applications. In particular, with the proliferation of GPS- and camera-enabled cell phones, a backup geolocation system is needed when GPS satellite signals are blocked in so-called "urban canyons." Image line segments are first located, and then the vanishing points of these segments are determined using the RANSAC robust estimation algorithm. Next, the intersections of line segments associated with pairs of vanishing points are used to generate local support for planar facades at different orientations. The plane support points are then clustered using an algorithm that requires no knowledge of the number of clusters or of their spatial proximity. Finally, building facades are identified by fitting vanishing point-aligned quadrilaterals to the clustered support points. Our experiments show good performance in a number of complex urban environments. The main contribution of our approach is its improved performance over existing approaches while placing no constraints on the facades in terms of their number or orientation, and minimal constraints on the length of the detected line segments.

Presented in this paper is the design of a skin filter which unlike many systems already developed for use, this system will not use RGB or HSI colour but an 8-bit greyscale instead. This is done in order to make the system more convenient to employ on an FPGA, to increase the speed to better enable real-time imaging and to make it easier to combine with the previously designed binary based algorithms. This paper will discuss the many approaches and methods that could be considered such as Bayes format and thresholds, pixel extraction, mathematical morphological strings, edge detection or a combination of the previous and a discussion about which provided the best performance. The research for this skin filter was carried out in two stages, firstly on people who had an ethnic origin of White - British, Asian or Asian British, Chinese and Mixed White and Asian. The second phase which won't be included here in great detail will cover the same principles for the other ethnic backgrounds of Black or Black British - Caribbean or Africa, Other Black background, Asian or Asian British - Indian, Pakistani or Bangladeshi. This is due to the fact that we have to modify the parameters that govern the detection process to account for greyscale changes in the skin tone, texture and intensity; however the same principles would still be applied for general detection and integration into the previous algorithm. The latter is discussed and what benefits it will give.

An adaptive skin segmentation algorithm robust to illumination changes and skin like backgrounds is presented in this paper. Skin pixel classification has been limited to only individual color spaces. There has not been a comprehensive evaluation of which color components or a combination of color components would provide the best skin pixel classification. Although the R, G, B components are the three primary features, transformation of these components to different color spaces provide additional set of features. The color components or the features present within a single color space may not be the best when it comes to skin pixel classification. In this paper an adaboost based skin segmentation technique is presented. Bayesian classifiers trained on the skin and non-skin probability densities specific color component spaces form the set of weak classifiers which adaboost is implemented. Additional classifiers are generated by varying the associated thresholds of the Bayesian classifiers. in An adaptive image enhancement technique is implemented to improve the illumination as well as the color of an image. This will enable to identify the skin pixels more accurately in the presence of non-uniform lighting conditions. Human skin texture is fairly uniform. This property is utilized to develop a method, which is based on the neighborhood information of a pixel. This step will provide more information in addition to color about a pixel being skin or non-skin. A comparison of the existing color based and neighborhood methods with the proposed technique is presented in this paper.

Detection and segmentation of objects of interest in image sequences is the first major processing step in visual surveillance applications. The outcome is used for further processing, such as object tracking, interpretation, and classification of objects and their trajectories. To speed up the algorithms for moving object detection, many applications use techniques such as frame rate reduction. However, temporal consistency is an important feature in the analysis of surveillance video, especially for tracking objects. Another technique is the downscaling of the images before analysis, after which the images are up-sampled to regain the original size. This method, however, increases the effect of false detections. We propose a different pre-processing step in which we use a checkerboard-like mask to decide which pixels to process. For each frame the mask is inverted to avoid that certain pixel positions are never analyzed. In a post-processing step we use spatial interpolation to predict the detection results for the pixels which were not analyzed. To evaluate our system we have combined it with a background subtraction technique based on a mixture of Gaussian models. Results show that the models do not get corrupted by using our mask and we can reduce the processing time with over 45% while achieving similar detection results as the conventional technique.

The purpose of the Automatic Target Recognition (ATR) Center is to develop an environment conducive to producing theoretical and practical advances in the field of ATR. This will be accomplished by fostering intellectual growth of ATR practitioners at all levels. From an initial focus on students and performance modeling, the Center's efforts are extending to professionals in government, academia, and industry. The ATR Center will advance the state of the art in ATR through collaboration between these researchers. To monitor how well the Center is achieving its goals, several tangible products have been identified: graduate student research, publicly available data and associated challenge problems, a wiki to capture the body of knowledge associated with ATR, development of stronger relationships with the users of ATR technology, development of a curriculum for ATR system development, and maintenance of documents that describe the state-of-the-art in ATR. This presentation and accompanying paper develop the motivation for the ATR Center, provide detail on the Center's products, describe the Center's business model, and highlight several new data sets and challenge problems. The "persistent and layered sensing" context and other technical themes in which this research is couched are also presented. Finally, and most importantly, we will discuss how industry, academia, and government can participate in this alliance and invite comments on the plans for the third phase of the Center.

Designing and testing algorithms to process hyperspectral imagery is a difficult process due to the sheer volume of the data that needs to be analyzed. It is not only time-consuming and memory-intensive, but also consumes a great amount of disk space and is difficult to track the results. We present a system that addresses these issues by storing all information in a centralized database, routing the processing of the data to compute servers, and presenting an intuitive interface for running experiments on multiple images with varying parameters.

In this paper, there is an investigation into the possibility of executing a Morphological Scene Change Detection (MSCD) system on a Field Programmable Gate Array (FPGA), which would allow its set up in virtually any location, with its purpose to detect intruders and raise an alarm to call security personal, and a signal to initial a lockdown of the local area. This paper will include how the system was scaled down from the full building multi-computer system, to an FPGA without losing any functionality using Altera's DSP Builder development tool. Also included is the analysis of the different situations which the system would encounter in the field, and their respective alarm triggering levels, these include indoors, outdoors, close-up, distance, high-brightness, low-light, bad weather, etc. The triggering mechanism is a pixel counter and threshold system, and its adaptive design will be included. All the results shown in this paper, will also be verified by MATLAB m-files running on a full desktop PC, to show that the results obtained from the FPGA based system are accurate.

This paper shows research into the development of techniques that can be used to recover what was written on a paper document after attempts have been made to obscure the content via methods such as burning or bleach for example. Here instead of using expensive high-tech imagery and infrared equipment, there is the aim of using off-the-shelf equipment to reduce economic costs in the form of a Sony Ericsson Mobile Phone with a 2.0 mega pixel camera with built in light. The latter was used in the data collection phase after the test documents were produced, various factors were considered here such as light reflection and incident angles on the paper, position of camera, light frequencies, visible light collection and night mode light collection in order to achieve the optimum test image. The FPGA was then brought in for the post-collection processing of the images using techniques currently developed using graphical block methodologies for ease of use, then the best string of operations to obtain the most efficient results of what was previously written will be presented by comparing it to a similar untouched document. The paper then explains the expansions to the experiments which include different types and coloured inks from various sources which include standard pens to inkjet printer cartridges on numerously coloured paper to see how truly effect the developed technique is.

In this paper, next generation network (NGN) based intermediate-view reconstruction (IVR) using variable block matching algorithm (VBMA) is proposed. In the proposed system, the stereoscopic images are estimated by VBMA, and they are transmitted to receiver through dynamic bandwidth allocation (DFA), this scheme improves a priority-based access network converting it to a flow-based access network with a new access mechanism and scheduling algorithm, and then 16-view images are synthesized by the IVR. From some experimental results, it is found that the proposed system improves peak signal-to-noise ratio (PSNR) up to 4.86 dB. Also, network service provider can provide upper limits of transmission delays by the flow. The modeling and simulation results with mathematical analyses obtained by this scheme are also provided.

In this paper, we provide a segmentation based Retinex for improving the visual quality of aerial images obtained under complex weather conditions. With the method, an aerial image will be segmented into different regions, and then an adaptive Gaussian based on the segmentations will be used to process it. The method addresses the problems existing in previously developed Retinex algorithms, such as halo artifacts and graying-out artifacts. The experimental result also shows evidence of its better effect.

In this work we propose a method for securing port facilities which uses a set of video cameras to automatically detect various vessel classes moving within buffer zones and off-limit areas. Vessels are detected by an edge-enhanced spatiotemporal optimal trade-off maximum average correlation height filter which is capable of discriminating between vessel classes while allowing for intra-class variability. Vessel detections are cross-referenced with e-NOAD data in order to verify the vessel's access to the port. Our approach does not require foreground/background modeling in order to detect vessels, and therefore it is effective in the presence of the class of dynamic backgrounds, such as moving water, which are prevalent in port facilities. Furthermore, our approach is computationally efficient, thus rendering it more suitable for real-time port surveillance systems. We evaluate our method on a dataset collected from various port locations which contains a wide range of vessel classes.

Statistical models of deformations are becoming crucial tools for a variety of computer vision applications such as regularization and validation of image registration and segmentation algorithms. In this article, we propose a new approach to effectively represent the statistical properties of high dimensional deformations. In particular, we propose techniques that use independent component analysis (ICA) in conjunction with wavelet packet decomposition. Two different architectures for ICA have been investigated; one treats the elastic deformations as random variables and the individual deformation field as outcomes and a second which treats the individual deformations as random variables and the elastic deformations as outcomes. The experiments were conducted using the Amsterdam Library of Images (ALOI), and the proposed algorithms were evaluated using the model generalization as a statistical measure. Experimental results show a significant improvement when compared to a recent deformation representation in the literature.