Search

OBJECTIVES Athletic performance and injury prevention are important for athletes and coaches. Different types of movement analyses have been created to aid in injury prevention and performance. The reliability of the Microsoft Kinect™ for movement analysis has not been widely tested. If reliable and accurate it could decrease the cost and time necessary for movement analysis. The purpose of this study was to determine if the Microsoft Kinect™ is an accurate measure of knee displacement during the parallel squat when compared to the Dartfish Team Pro Software 6.0. METHODS Twenty nine healthy recreational athletes participated in the study and used the Dartfish Team Pro Software 6.0 to validate the Microsoft Kinect™ as a tool to measure knee displacement. Subjects performed a parallel squat with a 2.1m long dowel rod. This exercise was used to compare value between systems. The intraclass correlation coefficient and paired-samples t-test were used to compare Dartfish Team Pro Software 6.0 and Microsoft Kinect™. Intrarater reliability of each system was also assessed. RESULTS There were 29 participants in the study. The interclass correlation coefficient for Dartfish Team Pro Software 6.0 and Microsoft Kinect™ showed that the Microsoft Kinect™ had a high-reliability ICC = 0.96. Intrarater reliability for Kinect™ and Dartfish were .98 and .99, respectively. The mean difference between systems for measured knee displacement was 1.06 cm. The mean for the Microsoft Kinect™ was 49.11 ± 1.9 and 50.16 ± 96 for the Dartfish (p > 0.05). CONCLUSIONS The Microsoft Kinect™ is reliable against the Dartfish Team Pro Software 6.0 as a tool to measure knee displacement using the parallel squat. It appears for healthy young adults, the Microsoft Kinect™ is reliable for movement analysis.

An over-sampling technique called V-synth is proposed and compared to borderline SMOTE (bSMOTE), a common methodology used to balance an imbalanced dataset for classification purposes. V-synth is a machine learning methodology that allows synthetic minority points to be generated based on the properties of a Voronoi diagram. A Voronoi diagram is a collection of geometric regions that encapsulate classifying points in such a way that any point within the region is closest to the encapsulated classifier than any other adjacent classifiers based on their distance from one another. Because of properties inherent to Voronoi diagrams, V-synth identifies exclusive regions of feature space where it is ideal to create synthetic minority samples. To test the generalization and application of V-synth, six databases from various problem domains were selected from the University of California Irvine's Machine Learning Repository. Though not always guaranteed due to the random nature of synthetic over-sampling, significant evidence is presented that supports the hypothesis that V-synth more consistently leads to the creation of more accurate and better-balanced classification models than bSMOTE when the classification complexity of a dataset is high.

Image based imposters suffer from common errors called disocclusion artifacts where portions of the scene that should be occluded by real geometry are visible when using image based imposters. These artifacts are the result of parallax error created by camera motion where regions of a mesh that were not visible at the time of imposter generation have become visible. This document presents an analysis of a computationally inexpensive on-line technique [Mourning et al. 2014] to resolve these disocclusions by stretching existing imposter [Maciel and Shirley 1995] texture information over new geometry bridging the gap between imposters. [Mourning et al. 2014] only presented automatic metrics showing improved image quality compared to traditional techniques; in order to corroborate the findings in [Mourning et al. 2014], human trials were performed to determine if human subjects found a similar increase in image quality. Results show a statistically significant improvement in image quality over traditional imposters.

This paper presents a technique for interactively colliding with and deforming mesostructures at a per-texel level. It is compatible with a broad range of existing mesostructure rendering techniques including both safe and unsafe ray-height field intersection algorithms. This technique is able to replace traditional 3D geometrical deformations (vertex-based) with 2D image space operations (pixel-based) that are parallelized on a GPU without CPU-GPU data shuffling and integrates well with existing physics engines. Additionally, surface and material properties may be specified at a per-texel level enabling a mesostructure to possess varying attributes intrinsic to its surface and collision behavior. Furthermore, this approach may replace traditional decals with image-based operations that naturally accumulate deformations without inserting any new geometry. This technique provides a simple and efficient way to make almost every surface in a virtual world responsive to user actions and events. It requires no preprocessing time and storage requirements of one additional texture or less. The algorithm uses existing inverse displacement map algorithms as well as existing physics engines and can be easily incorporated into new or existing game pipelines.

Detecting and recognizing objects in environments with unpredictable illumination changes remains a challenging task. Existing algorithms employ a passive methodology to deal with these environments, where learning is performed from many samples taken under various lighting conditions or with some pre-designed color constancy models. In this paper, the challenges of unpredictable illumination changes are addressed through a feedback strategy. With the use of feedback, self-adaptation in object detection and recognition is possible in response to variable illumination. Self-adaptation is achieved through feedback from the recognition phase to the detection phase. A multilevel Markov random field (MRF) is adopted to model both the detection and recognition processes. The original MRF approach is extended to a model that encodes simultaneous object detection and recognition. Experimental results show the feasibility of the proposed framework.

In this paper, a novel machine vision application is presented for analyzing and visualizing confocal microscopy images of biological preparations. The proposed system is divided into three subsystems: a 3D curved surface extraction subsystem that generates 3D surfaces passing through selected key points in confocal image stacks; a 2D image projection subsystem that produces a flattened projection of the extracted curved surface; and an image mosaic subsystem that concatenates a series of image projections to form a view of an entire biological preparation. A combination of cubic interpolation and boundary matching is employed to reconstruct the 3D curved surface that passes through selected key points. The projection process integrates data fidelity and local smoothness constraints, producing a color or intensity projection along the desired 3D surface. Registration is achieved by aligning and minimizing the sum of the squared distances (SSD) between the intensities of the corresponding pixels. Two biological applications of the proposed system are reported to illustrate how the vision system could aid in biological research.

For computer vision systems to operate in many real-world environments, processing must occur in real-time under dynamic conditions. An agent-based methodology offers an approach to increase flexibility and scalability to accommodate the demands of a real-time, dynamic environment. This paper presents an agent-based architecture that uses a utility optimization technique to guarantee that important vision tasks are fulfilled even under resource constraints. To ensure that the processing of vision tasks is both reliable and flexible, multiple behaviors are utilized to accomplish the vision application's requirements. A vision behavior consists of a grouping of vision algorithms and a set of service levels associated with these algorithms. Utility functions are adopted to evaluate the performance of all possible behaviors that can address the requirements of a vision application within resource constraints. The maximum overall utility corresponds to the optimal behavior. Two example systems using this model are presented to show the applicability of the architecture. Experimental results show that this agent-based architecture outperforms traditional non-agent-based approaches.

Image based imposters suffer from common errors called disocclusion artifacts where portions of the scene that should be occluded by real geometry are visible when using image based imposters. These artifacts are the result of parallax error created by camera motion where regions of a mesh that were not visible at the time of imposter generation have become visible. This paper presents a computationally inexpensive on-line technique to resolve these disocclusions by stretching existing imposter texture information over new geometry bridging the gap between imposters. This paper also includes an analysis of six automatic metrics for similarity between output frames of a flight path using this novel technique and the impersonated data, to quantify the quality of the technique compared to the original scene and other level-of-detail techniques.

A similarity structure (shape) of a 3-D object can be reconstructed from two perspective views of the object obtained by a calibrated camera. One of the best-known methods for performing such a reconstruction is the 8-point algorithm. To evaluate the psychological plausibility of this algorithm, we compared its performance to that of human subjects in a task of binocular shape reconstruction. Results show that the reconstructions produced by the 8-point algorithm are substantially less accurate than those produced by human subjects. We then modified the 8-point algorithm by incorporating natural constraints that are present in human binocular vision. Specifically, the new algorithm simulates a fixating system with no torsion. The performance of this algorithm is more robust in the presence of noise compared to that of the 8-point algorithm and is similar to the performance of human subjects.

Parallel processing methods are a means to achieve significant speedup of computationally expensive image understanding algorithms, such as those applied to range images. Practical implementations of these algorithms must deal with the problems of selecting an appropriate parallel architecture and mapping the algorithm onto that architecture. The parallel implementation approaches for range image segmentation that are presented here are applicable to many low-level image understanding algorithms in a variety of parallel architectures. An evaluation of initial data distribution is presented to determine whether a square subimage or a striped subimage distribution would result in the greatest overall reduction in execution time for the given range image segmentation problem. Novel implementations that consider each data distribution's treatment of edge pixels in window operations yield a trade-off between the number of data transfers versus the amount of computation. This trade-off is examined both analytically and experimentally. Additionally, using the same initial data distributions, a technique is introduced for changing the allocation of work to each of the processors to reduce the number of network settings by one half. This technique and the method for determining the better initial data distribution can be used with any machine and any window-based technique that requires a full window to perform image calculations. Comparisons of range image processing algorithms are performed using “pure” SIMD algorithms, “pure” MIMD algorithms, and mixed-mode implementations with both SIMD and MIMD elements. Each of these approaches are quantitatively analyzed and compared for implementing the different phases of a particular hybrid range segmentation algorithm. Results of this implementation study indicate that quantifiable reductions in execution time result from the proper choice of parallel mode for each portion of the segmentation process.

This research features the rapid recognition of three-dimensional objects, focusing on efficient indexing. A major concern in practical vision systems is how to retrieve the best matched models without exploring all possible object matches. We have employed a Bayesian framework to achieve efficient indexing of model objects. A decision-theoretic measure of the discriminatory power of a feature for a model object is defined in terms of posterior probability. Domain-specific knowledge compiled off-line from CAD model data is used in order to estimate posterior probabilities that define the discriminatory power of features for model objects. In order to speed up the indexing or selection of correct objects, we generate and verify the object hypotheses for features detected in a scene in the order of the discriminatory power of these features for model objects. Based on the principles described above, we have implemented a working prototype vision system using a feature structure called an LSG (local surface group) for generating object hypotheses. Our object recognition system can employ a wide class of features for generation of object hypotheses. In order to verify an object hypothesis, we estimate the view of the hypothesized model object and render the model object for the computed view. The object hypothesis is then verified by finding additional features in the scene that match those present in the rendered image. Experimental results on synthetic and real range images show the effectiveness of the indexing scheme.

In this paper, we propose the volume between two surfaces normalized by the surface area (interpreted as average distance between two surfaces) as an invariant quantitative measure for comparing surface reconstruction results of the explicit form, z(x,y). The invariant property of the volume quantity provides the same measure with respect to an arbitrary coordinate system. By normalizing the volume by the surface area, the values of the measure can be compared for different size of images. We also present a novel computationally simple and efficient way of computing the volume between two surfaces and the surface area using a least-squared-error plane approximation of a surface patch defined over a rectangular grid. Experiments indicate that the method gives equivalent performance as other more complicated and computationally expensive methods. The advantages of this new method are that computation is simple and efficient.

This paper presents a technique for interactively deforming and colliding with mesostructures at a per-texel level. It is compatible with a broad range of existing mesostructure rendering techniques including both safe and unsafe ray-height field intersection algorithms. This technique integrates well with existing physics engines and is able to reduce traditional 3D geometrical deformations (vertex-based) to 2D image space operations (pixel-based) that are parallelized on a GPU without CPU-GPU data shuffling. Additionally, surface and material properties may be specified at a per-texel level enabling a mesostructure to possess varying attributes intrinsic to its surface and collision behavior; furthermore, this offers an image-based alternative to traditional decals. This technique provides a simple way to make almost every surface in a virtual world responsive to user actions and events. It requires no preprocessing time and storage requirements of one additional texture or less. The algorithm uses existing displacement map algorithms as well as existing physics engines and can be easily incorporated into new or existing game pipelines.

Recently, there has been a great increase in interest in using three dimensional stereoscopic displays to provide viewers with realistic 3D views of objects of interest. Some applications where stereoscopic displays are becoming popular include medical visualization, visualization of meteorological data, and various virtual reality applications. To quantify the effectiveness of stereoscopic systems over conventional monoscopic systems, well-designed experiments and data analysis methods are necessary. This task requires the combined effort of application scientists and experts in experimental design. Lack of interdisciplinary collaboration is a primary weakness of many stereoscopic display studies, resulting in the neglect of many important but subtle experimental issues. In this paper, we discuss specific issues that arise in the design of studies to determine the effectiveness of digital stereo imagery. Issues concerning statistical analysis of the experimental data are also discussed. References to related literature from engineering, computer graphics, and psychophysics are given. The issues developed herein provide a guideline for the design of studies to compare observer performance when using different imaging modalities.

This paper describes the application of a first order regularization technique to the reconstruction of visible surfaces. Our approach is a computationally efficient first order method that simultaneously achieves approximate invariance and preservation of discontinuities. It is also robust with respect to the smoothing parameter /spl lambda/. The robustness property to /spl lambda/ allows a free choice of /spl lambda/ without struggling to determine an optimal /spl lambda/ that provides the best reconstruction. A new approximately invariant first order stabilizing function for surface reconstruction is obtained by employing a first order Taylor expansion of a nonconvex invariant stabilizing function that is expanded at the estimated value of the squared gradient instead of at zero. The data compatibility measure is the squared perpendicular distance between the reconstructed surface and the constraint surface. This combination of stabilizing function and data compatibility measure is necessary to achieve invariance with respect to rotations and translations. Sharp preservation of discontinuities is achieved by a weighted sum of adjacent pixels. The results indicate that the proposed methods perform well on sparse noisy range data. In addition, the volume between two surfaces normalized by the surface area (interpreted as average distance between two surfaces) is proposed as an invariant measure for the comparison of reconstruction results. >

This paper describes a visualization system for the Joint Precision Aircraft Landing System (JPALS) that integrates remote sensor data with well-known databases of terrain imagery, elevation, and surveyed aviation locations. This visualization system allows teams of research engineers and managers to collaborate in real-time, during test flights, in a distributed fashion. Each user interacts with their own copy of the virtual world, updated constantly in real-time, to manipulate and investigate in a fashion that best accomplishes their own specific task. This allows the entire team to collaborate simultaneously, which mitigates the common latency associated with team members performing offline tasks, such as inspecting log files or verifying MatLab plots. Test flights utilizing our visualization system have shown that it enabled the research team to make decisions on the fly, such as changing the antenna type and placement, and also allowed post-flight analysis and prediction of desirable changes. It could, for instance, predict optimal antenna placement for a given flight path. This can save a project the large expense of performing additional test flights if analysis had been performed offline between flights. Since our system is GPS based, it can be used as an alternate truth system to validate traditional instrument approach and landing systems. The recording and playback features of our visualization can also be used as a teaching aid in flight training for both pilots and air traffic controllers.

A new algorithm for ray tracing generalized cylinders whose axis is an arbitrary three-dimensional space curve and whose cross-sectional contour can be varied according to a general sweeping rule is presented. The main restriction placed on the class of generalized cylinders that can be ray-traced is that the sweeping rule of the generalized cylinder must be invertible. This algorithm handles a broader class of generalized cylinders than any other reported ray tracer. It has been integrated into a general geometric modeling system that can render objects utilizing visible light as well as simulated X rays. Generalized cylinders are often used in modeling systems because they compactly represent objects. Many commonly occurring objects including snakes, horses, airplanes, flower vases, and organs of the human abdomen such as the stomach and liver can be described naturally and conveniently in terms of one or more generalized cylinder primitives. By extending the class of generalized cylinders that can be conveniently modeled, the presented algorithm enhances the utility of modeling systems based on generalized cylinders. X-ray images of the internal bone structure of a knee joint and a visible light image of a fan blade assembly are presented.

Virtual environments are a topic of discussion for many in the business and commerce fields. However, K-12 school systems have been slow to embrace technology within the typical classroom environment, much less a virtual environment. This chapter examines the development of Web-based Flash® content, an Interactive Science Lab in Second Life®, and newly created software called the STEAMiE engine being used by rural Appalachian middle school science classrooms in Ohio, USA. Presented are discussions of the design process through the use of teachers as content experts, the designing of games and simulations for middle school children, the changes in teaching strategies of middle school science teachers, and Fellows and the expectations of school administrators, teachers, and students during design and implementation.

A flying aircraft disturbs the local atmosphere through which it flies creating a turbulent vortex at each wing tip known as a wake vortex. These vortices can persist for several minutes and endanger other aircraft traversing that turbulent airspace; large vortices are essentially invisible horizontal tornadoes and are a grave threat to smaller aircraft, especially during landing and take off. Accidents related to wake turbulence have resulted in both loss of life and aircraft destruction in the United States and around the world. Currently no cockpit instrumentation exists that tracks wake vortices and enables a pilot to sense and avoid wake turbulence in real-time. This paper presents a prototype of a novel, flight tested instrument that tracks wake vortices and presents this information to a pilot in real time using a synthetic virtual world augmented with wake turbulence information.

LIDAR (Light Detection and Ranging) [1] is a remote sensing technology that is growing in popularity in varied and diverse disciplines. Modern LIDAR systems can produce substantial amounts of data in very brief amounts of time, so one of the greatest challenges facing researchers is processing and visualizing all of this information, particularly in real time. Ideally, a scientific visualization of a set of LIDAR data should provide an accurate view of all the available information; however, sometimes it is beneficial to exchange a small portion of that accuracy for the increased usability and flexibility of a real-time interactive display. The goal of this research is to characterize under what conditions the level-of-detail rendering technique known as impostors [2-4] can effectively optimize the inherent trade-offs between accuracy and interactivity in large-scale point cloud datasets.

Robust Point Matching (RPM) is a common image registration algorithm, yet its large computational complexity prohibits registering large point sets in a timely manner. With recent advances in General Purpose Graphical Processing Units (GPGPUs), commodity hardware is capable of greatly reducing the execution time of RPM when non-rigidly aligning thousands of data points. In this paper, we identify areas where parallelism can be exploited in the RPM algorithm, and investigate a GPU-based approach to accelerate the implementation. Other common RPM implementations are compared with our solution. Experiments on synthetic and real data sets show that our approach achieves close to linear speed-up with respect to total computational power over the widely used Matlab implementation. Our tests indicate that utilizing our implementation on current state of the art GPU technology would enable the use of vastly greater point set sizes.

Today, there exists a significant disparity in the degree of technological development between commercial and educational games. The STEAMiE educational engine is a cutting-edge system used to create advanced, realistic, immersive learning environments. These environments have been shown to influence comprehension and retention of hard-to-teach science concepts. This paper will present an overview of the features of our STEAMiE educational engine and how to use it to create games/simulations capable of enhancing a userpsilas learning experience. The STEAMiE educational engine contains a rich feature set allowing for development of powerful modules within a short period of time. From a development point of view, the STEAMiE educational enginepsilas object oriented design is modular and can be easily extended to support new functionality in a timely manner. This significantly shortens development time and abstracts complexities from the developer allowing for clean and efficient implementations of games and simulations.

This paper discusses a case study which aims to make research projects more engaging for middle school students by incorporating a variation to the traditional gallery walk workflow. The process begins by having the students conduct individual research on one of a finite list of topics and produce a poster that details what they have learned. We collect the research posters and transfer the media into the shared virtual world of second life, where a developer organizes the posters into a ldquovirtual museumrdquo of full-color high-quality panels displaying the studentspsila creations. Students return to the computer lab, log in to the virtual world, and explore the virtual gallery. This gives students the opportunity to learn more about the other topics from their peers in an engaging, interactive, and fun way. We evaluate the effectiveness of this workflow by injecting quizzes on the topic material before the research phase, after the research phase, and after using the virtual gallery walk. As virtual worlds are rarely utilized in K-12 classrooms, the virtual gallery walk is innovative as a vehicle for student learning and as a possible outlet for sharing the work with community of other students active within second life.

Automatic segmentation of sub-cortical structures has great use in studying various neurodegentative diseases. In this paper, we propose a fully automatic solution to this problem through the utilization of a distribution atlas built from a set of training MR images. Our model consists of two major components: a local likelihood based active contour (LLAC) model and a guiding probabilistic atlas. The former has a very strong ability in standing out the structures that are in low contrast with the surrounding tissues. The latter has the functionality of defining and leading the segmentation procedure to capture the structure of interest. Formulated under the maximum a posterior framework, probabilistic atlas for the structure of interest, e.g. caudate, putamen, can be seamlessly integrated into the level set evolution procedure, and no thresholding step is needed for capturing the target.

This work-in-progress paper describes a web- based system that is being developed at Ohio University for developing and administering active learning assignments. The paper describes the design of the system, and its intended use in outcomes-based assessment of student learning in Computer Science. The system allows students to access active learning exercises via the web and provides students with immediate feedback on their work. Faculty use the system to design active learning assignments that tie student performance to achievement of specific course outcomes. This work-in-progress paper describes initial experiences with this system in an advanced course where both the software package JFLAP (An Interactive Formal Languages and Automata Package) and C++ programming are used.

In this paper, we propose a distribution-based active contour model for brain MRI segmentation. As a generalization of the Chan-Vese piecewise-constant model, our solution uses Bayesian a posterior probabilities as the driving forces for curve evolution. Distribution prior, if available, can be seamlessly integrated into the level set evolution procedure. Unlike other region-based active contour models, our solution relaxes the global piecewise-constant assumption, and uses locally varying Gaussians to better account for intensity inhomogeneity and local variations existing in many MR images. More accurate and robust segmentations are therefore achieved. Experiments conducted on synthetic and real brain MRIs demonstrate the improvement made by our model.

Detecting and recognizing objects in unstructured environments is one of the most challenging tasks in computer vision research. We propose an innovative algorithm, called deformable illumination, to address the problem of illumination variance in natural environments. Parallel to the role of deformable shape in object recognition, deformable illumination is designed as an object detection technique. A unified framework presented here integrates both deformable illumination and deformable shape as a simultaneous scheme for object detection and recognition in unstructured environments. Experimental results show the effectiveness of deformable illumination in addressing illumination variance.

This paper describes a test-bed for technology that unifies agent based computing and adaptive resource management for dynamic real-time systems. We describe a unified framework that combines a hybrid agent based architecture with explicit resource adapting mechanisms.

Facial event mining is one of the key techniques for automatic human face analysis. It plays an important role in human computer interaction. This paper proposes a new approach to facial event recognition by combining active shape models (ASMs) and coupled hidden Markov models (CHMMs). Based on the assumption that a complex facial event can be decomposed into multiple coupled processes, ASMs are used to track global facial features and to decouple pattern attributes for upper and lower faces separately. These two interacting processes are modeled as a CHMM for training and recognition. Four basic facial events are investigated. Preliminary experiments yield consistent results that show the significant advantage of CHMMs over conventional HMMs for facial event mining in video.

As one of the key techniques for futuristic man-machine interface, facial expression analysis has received much attention in recent years. This paper proposes a hierarchical approach to facial expression recognition in image sequences by exploiting both spatial and temporal characteristics within the framework of hierarchical hidden Markov models (HHMMs). Human faces are automatically detected in the maximum likelihood sense. Gabor-wavelet based features are extracted from image sequences to capture the subtle changes of facial expressions. Four prototype emotions; i.e. happiness, anger, fear and sadness, are investigated using the Cohn-Kanade database and an average of 90.98% person-independent recognition rate is achieved. We also demonstrate that HHMMs outperform HMMs for modeling image sequences with multilevel statistical structure.

A solution to the problem of model-based image segmentation by integrating prior color and texture information into a parametric active contour model is proposed in this paper. The vector flow field technique is adopted as the computational scheme because of its large effective capture range. Color and texture priors are modeled in a uniform way to generate a potential field and its associated vector flow field for the repetitive application of the gradient vector field (GVF). The initialization of the snake is decided by studying the shape properties of the potential field. The shape information can be utilized to avoid unnecessary initializations caused by disturbing background similar color and texture features to the model object. The potential field is also valuable for determining the weights for integrating the color and texture based vector flow field and the edge based vector flow field. With the proper integration weights, the snake evolved on the combined vector flow field can bypass strong edges in the background and stop correctly on the object's boundary. Experimental results are shown on real scenes with complex backgrounds.

Although significant research effort has been devoted to the natural scene analysis, there hasn't been much work on natural scene synthesis. The research of such nature can help to understand the hidden structure of natural texture and the illumination patterns for classification and segmentation purposes. The common texture synthesis methods are not suitable for natural scene synthesis due to inhomogeneity of the natural textures. In this paper, we propose a new natural scene synthesis framework based on multiple eigenspace decomposition, which is a natural extension of principle component analysis (PCA). Two levels of multiple eigenspaces are used; label level multiple eigenspaces for generating depth map and color level multiple eigenspaces for generating nature textures to fill the label maps. Experimental results are presented to show the effectiveness of the proposed framework.

Strong image segmentation is a very challenging problem in computer vision research. Both data-driven and model-driven approaches have been investigated in the past two decades, and many approaches proposed. Although model-based approaches are more promising in addressing strong image segmentation, data-driven approaches present more general frameworks which could potentially be adopted to segment general scenes without any prior model information. We discuss the problems of strong image segmentation from a data-driven perspective, and present a modeling technique describing an object with both its segments and a hierarchical relationship among the segments. The paper is devoted to the discussion of the feasibility of data-driven approaches for strong image segmentation. Existing approaches are not suitable for strong image segmentation in complex environments, but preliminary experimental results show the feasibility of our proposed model.

The paper presents a model that is useful for developing resource allocation algorithms for distributed real-time systems that operate in dynamic environments. Interesting aspects of the model include dynamic environments, utility and service levels, which provide a means for graceful degradation in resource-constrained situations and support optimization of the allocation of resources. The paper also provides an allocation algorithm that illustrates how to use the model for producing feasible, optimal resource allocations.

For accurate segmentation of natural scenes, there is a need for powerful, yet simple, color and texture models. Many of the existing segmentation techniques are able to segment homogeneous color scenes as well as the textured images. Despite their computational power, these methods require the estimation of model parameters that are usually difficult to obtain. We propose a simple, yet effective, image model that can efficiently describe natural textures for color scene segmentation and, in turn, be utilized in texture synthesis.

Parallel processing methods are an attractive means to achieve significant speedup of computationally expensive image understanding algorithms, such as those applied to range images. Mixed-mode parallel systems are ideally suited to this area because of the flexibility in using the different modes of parallelism. The trade-offs of using different parallel modes are examined through the implementation of hybrid range segmentation operations, characteristic of a broad class of low level image processing algorithms. Alternative means of distributing data among the processing elements that achieve improved performance are considered. Results comparing different implementations on a single reconfigurable parallel processor. PASM, indicate some generally applicable guidelines for the effective parallelization of vision algorithms. >

The authors discuss the use of stereo imaging techniques as a low cost and low radiation dosage means to provide physicians with three-dimensional (3D) information for diagnosis and follow-up care of patients. Instead of using a pair of images obtained from perpendicular directions (as in the case of biplane radiography), the authors show that if a pair of small-angle stereo X-ray images are obtained, observers can vividly perceive the images in 3D during acquisition, i.e. real-time stereo visualization is possible. In addition, quantitative information such as blood flow measurements and the amount of arterial narrowing can be extracted off-line by using computational stereo algorithms. The potential advantages of using a small-angle stereo-based approach as opposed to biplane angiography are discussed. >

A design for a system to perform object recognition from a large model data base is presented, focusing on efficient indexing. The authors propose a decision-theoretic approach using a Bayesian framework to achieve efficient indexing of model objects. A decision-theoretic measure of the discriminatory power of a feature for a model object is defined in terms of posterior probability. Domain-specific knowledge compiled off-line from CAD model data is used in order to estimate posterior probabilities that define the discriminatory power of features for model objects. Detectability of a feature defined as a function of the feature itself, viewpoint, sensor characteristics, and the feature detection algorithm(s) is also considered in the computation of discriminatory power. In order to speed up the indexing or selection of the correct objects, the authors generate and verify the object hypotheses for the features detected in the scene in the order of the discriminatory power of these features for model objects. >

Traditional methods for computing optical flow are mainly based on image brightness constancy. In the real world the brightness constancy usually does not hold. Here we present the idea of using invariant feature based on the brightness change model to estimate the optical flow. Both the mathematical derivation and the experiments show that the new model is better than brightness based optical flow constraint.

Agent based computing offers the ability to decentralize computing solutions by incorporating autonomy and intelligence into cooperating, distributed applications. It provides an effective medium for expressing solutions to problems that involve interaction with real-world environments and allows modelling of the world state and its dynamics. This model can be then used to determine how candidate actions affect the world and how to choose the best from a set of actions. Most agent paradigms overlook real-time requirements and computing resource constraints. We discuss the application of agent based computing to RoboCup and examine methods to improve it. In particular, we discuss the incorporation of a meta-level reasoning mechanism that handles individual agent organization, plan generation, task allocation, integration and plan execution. We also propose an architecture where a meta-agent is further enhanced by combining it with system-level resource allocation and optimization. The approach adopted by us unifies agent based computing with adaptive resource management for dynamic real-time systems. The goal is to build and implement a distributed, intelligent, agent based system for dynamic real-time applications.

Breakthrough in Earth Science Observing will occur when constellations of Earth observing satellites are able to fully collaborate together and collectively monitor the conditions of our planet through a vast array of instruments. These satellites form a network that consists of distributed processes that need to respond to perceived scientific events, the spacecraft environment, spacecraft anomalies and user commands. The requests and responses exhibit dynamic behavior. In order to handle such dynamic environments, a method is needed to guarantee the real-time quality of service constraints. The DeSiDeRaTa resource management approach is being enhanced to characterize the dynamic aspects of intraconstellation topologies and to accommodate the concept of service levels and utility.This paper presents a design model of cooperative problem solving to show how the solution approach addresses the key challenges presented in the problem and specifies how the agent, resource manager and satellite constellations would operate correctly and interact in complex, dynamic and unpredictable environments. It extends the system model of DeSiDeRaTa to accommodate the concepts of utility, service levels and planning. The system model for the IPA is presented to show the proof of concept.

Reports the diagnostic performance of observers in detecting abnormalities in computer-generated mammogram-like images. A mathematical model of the human breast is defined in which breast tissues are simulated by spheres of different sizes and densities. Images are generated by casting rays from a specified source, through the model, and onto an image plane. Observer performance when using two viewing modalities (stereo versus mono) is compared. In the stereo viewing mode, images are presented to the observer (wearing liquid-crystal display glasses), such that the left eye sees the left image only and the right eye sees the right image only. In this way, the images can be fused by the observer to obtain a sense of depth. In the mono viewing mode, identical images are presented to the left and right eyes so that no binocular disparities will be produced by the images. Observer response data are evaluated using receiver operating characteristic (ROC) analysis to characterize any difference in detectability of abnormalities (in either the density or the arrangement of simulated tissue densities) using the two viewing modes. The authors' experimental results indicate the clear superiority of stereo viewing for detection of arrangement abnormalities. For detection of density abnormalities, the performance of the two viewing modes is similar. These preliminary results suggest that stereomammography may permit easier detection of certain tissue abnormalities, perhaps providing a route to earlier tumor detection in cases of breast cancer. >

To determine the effectiveness of stereo imaging in aiding the detection of objects in a scene, we are conducting experiments in which subjects are shown computer-generated stereo and mono images and are asked to determine if there is an object with particular characteristics in the image. The experimental data are analyzed using receiver operating characteristic (ROC) approaches to determine which types of objects may be easier to detect using stereo viewing. In this paper, issues that rise in the design of ROC studies to determine the statistical effectiveness of stereo imagery are discussed. These include traps and pitfalls such as varying viewing conditions, image intensity differences, ghosting, flicker, the speed/accuracy tradeoff, subjects' stereo acuity, and degree of difficulty in the discrimination task. Our experimental results show that when these problems are properly addressed, stereo viewing increases the sensitivity and specificity of observer performance in detecting subtle features in simulated x- ray transmission images.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This paper describes the application of regularization techniques to the problem of segmenting range images. We propose a new energy functional that varies the amount of smoothing according to the gradient of the data. An iterative application of reconstruction using this new functional improves the signal/noise ratio of the noisy input image with good preservation of discontinuities. By employing reconstruction using this new energy functional, the difficulty in applying regularization techniques to the segmentation problem due to smoothing over discontinuities is circumvented. The results indicate that the algorithm performs especially well on noisy range images. Reconstruction using the new energy functional shows the possibility of its application to the problem of image enhancement. An algorithm is described for the detection of zeroth order discontinuities and surface reconstruction. We also discuss how the same algorithm can be applied to detect first order discontinuities and be applied to gradient reconstruction.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This paper discusses an ongoing development effort to utilize knowledge based techniques for analyzing terrain. The application focuses on analyzing terrain to determine the likely location of various military forces and to determine the movement (traversability) and likely classifications of those forces. The effort has involved (i) the development of representations of terrain objects that facilitate automated reasoning and (ii) the development of a rule based system that reasons about the terrain objects and develops hypotheses about terrain characteristics. The system is largely goal directed and can answer questions such as: Is this a good location for a headquarters unit? Is there a traversable path (by, say, a tank) between points A and B? Where are likely locations for artillery sites? The rules from the system were obtained from domain experts on military tactics. The terrain representation is semantically oriented (rather than grid oriented). The scene is represented as a series of "trees" (using largely quadtree and k-d tree formulations) with each tree representing a terrain feature. Trees are logically "combined" to fire more complex rules. The data in the system can be derived from DMA data bases or other sources.

The paper proposes an adaptive framework for shape-based image retrieval with relevance feedback. The motivation is to find an adjustable shape representation scheme that can account for feedback information. Relevance feedback is modeled as a dynamic eigenspace decomposition, and is used to classify the database into relevant and irrelevant groups with respect to the query. Eigenvectors of the subspace are updated by optimizing a linear transform with respect to the J3 class separability criterion. Experimental results show that the proposed approach can effectively capture a user's perceptual subjectivity