The need to generate a spider to provoke a desired anxiety response arises in the context of personalized virtual reality exposure therapy (VRET), a treatment approach for arachnophobia. This treatment involves patients observing virtual spiders in order to become desensitized and decrease their phobia, which requires that the spiders elicit specific anxiety responses. However, VRET approaches tend to require therapists to hand-select the appropriate spider for each patient, which is a time-consuming process and takes significant technical knowledge and patient insight. While automated methods exist, they tend to employ rules-based approaches with minimal ability to adapt to specific users. To address these challenges, we present a framework for VRET utilizing procedural content generation (PCG) and reinforcement learning (RL), which automatically adapts a spider to elicit a desired anxiety response. We demonstrate the superior performance of this system compared to a more common rules-based VRET method.

Optimizing the configuration and placement of sensors is crucial for reliable fall detection, indoor localization, and activity recognition in assisted living spaces. We propose a novel, sample-efficient approach to find a high-quality sensor placement in an arbitrary indoor space based on grey-box Bayesian optimization and simulation-based evaluation. Our key technical contribution lies in capturing domain-specific knowledge about the spatial distribution of activities and incorporating it into the iterative selection of query points in Bayesian optimization. Considering two simulated indoor environments and a real-world dataset containing human activities and sensor triggers, we show that our proposed method performs better compared to state-of-the-art black-box optimization techniques in identifying high-quality sensor placements, leading to an accurate activity recognition model in terms of F1-score, while also requiring a significantly lower (51.3% on average) number of expensive function queries.

This paper explores the competency of the time domain ultra-wideband (UWB)-circular synthetic aperture radar (CSAR) to image the breast and detect tumors. The image reconstruction is performed using a time domain global back projection technique adapted to the circular trajectory data acquisition. This paper also proposes a sectional image reconstruction method to compensate for the group velocity changes in different layers of a multilayer medium. Experiments on an advanced breast phantom examines the suitability of this technique for breast tumor imaging. The advanced breast phantom is designed based on a MRI of a real patient, fabricated using 3D printing technology, and filled with liquids that emulate normal and cancerous tissues. The measurement results, compared with MRI imaging of the phantom, demonstrate the suitability of the UWB-CSAR method for breast tumor imaging. This method can be a tool for early diagnosis as well as for treatment monitoring during chemotherapy or radiotherapy.

It is critical to know what brain regions are affected by an ischemic stroke, as this enables doctors to make more effective decisions about stroke patient therapy. These regions are often identified by segmenting computed tomography perfusion (CTP) images. Previously, this task has been done manually by an expert. However, manual segmentation is an extremely tedious and time-consuming process, that is not suitable for ischemic stroke lesion segmentation as it is highly time sensitive. In addition, these approaches require an expert to do the segmentation task, who may not be available and are prone to errors. Several automatic medical image analysis methods have been proposed for ischemic stroke lesion segmentation. These approaches, typically, use hand-crafted features that are predefined to represent the input data. However, because of the irregular and physiologically shapes, ischemic stroke lesions cannot be properly predicted, in an automatic way, using simple predefined features. In this work, we propose an automatic prediction algorithm that learns an effective model for segmenting the ischemic stroke lesion. This learned model first uses four 2D U-Nets to, separately, extract valuable information about the location of the stroke lesion from four CTP maps (CBV, CBF, MTT, Tmax). The model then combines the probability maps extracted by the U-Nets, to decide whether the pixels are either lesion or healthy tissues. This approach uses information about each pixel, as well as its neighborhood, to learn the stroke lesion, despite their varying shapes. The segmentation performance is evaluated using dice similarity coefficient (DSC), volume similarity (VS), and Recall. We have used this new algorithm on ISLES 2018 challenge dataset and found that our approach achieved results that are better than state-of-the-art approaches.

We propose a method for estimating blood pressure (BP) non-invasively from electrocardiogram (ECG) and photoplethysmogram (PPG) signals. This method has potential to be used as a continuous form of BP estimation. Along with these signals, to our knowledge, for the first time in the BP measurement studies, we included accelerometric and gyroscopic signals from a wearable device to compensate for motion during continuous BP prediction. Our prediction model is a long-short-term-memory (LSTM) architecture of a recurrent neural network (RNN), which accommodates the multiscale temporal dependency between the sequential raw signal values and the corresponding systolic and diastolic BP values. We performed a study with 50 healthy volunteers. The mean difference ± standard deviation (SD) of the RNN-based approach were 0.02±4.8 for SBP and 1.5±3.7 for DBP in seated position & 2.6±6.0 for SBP and 2.7±4.5 for DBP while walking. These values meet current validation standard requirements for measurement accuracy. Our experiments also demonstrate that the proposed RNN-based approach outperformed the classical linear regression model for BP prediction.

In this paper, we propose an automated segmentation algorithm for human leg muscles from 3D MRI data using a deep convolutional neural network (CNN). Using a generalized cylinder model, a 3D human leg muscle is represented by two smooth 2D images. The CNN predicts these two images from raw 3D voxels. For our base CNN, we use a pre-trained AlexNet that is coupled with a principle components analysis (PCA)-head. The AlexNet predicts a compressed vector, which is then back-projected by the PCA-head into two 2D images representing a 3D leg muscle boundary. This structured-output CNN architecture is fine-tuned in an end-to-end fashion. Our proposed CNN outperforms the conventional model-based approach, the active appearance model (AAM) image segmentation algorithm. The average Dice score between the ground truth segmentation and the obtained segmentation image is 0.85 using our CNN model, whereas the AAM yields a Dice score of 0.60.

The internal anatomy of a patient can be difficult for clinicians and other patients to visualize and analyzed in context. The ability to project medical images (CT and MRI scans) directly onto a patient's body helps those viewing these images to recover missing anatomical context for more accurate interpretation. This paper proposes such a system. Various types of images can be displayed using volume rendering techniques for realistic visualization of the internal anatomy and 3D models from segmented images. Calibration is performed on multiple systems to obtain an accurate common coordinate system, as well as correcting visual distortions from the cameras and the projector. This projected AR system provides a common perspective that is not tied to an individual point-of-view which can be used by others such as a surgical team. The system is easily extendable to other display technology and has many potential applications including education, surgical planning, laparoscopic surgery, and entertainment.

Previous image parsing methods usually model the problem in a conditional random field which describes a statistical model learned from a training dataset and then processes a query image using the conditional probability. However, for clothing images, fashion items have a large variety of layering and configuration, and it is hard to learn a certain statistical model of features that apply to general cases. In this article, we take fashion images as an example to show how Markov Random Fields (MRFs) can outperform Conditional Random Fields when the application does not follow a certain statistical model learned from the training data set. We propose a new method for automatically parsing fashion images in high processing efficiency with significantly less training time by applying a modification of MRFs, named reweighted MRF (RW-MRF), which resolves the problem of over smoothing infrequent labels. We further enhance RW-MRF with occlusion prior and background prior to resolve two other common problems in clothing parsing, occlusion, and background spill. Our experimental results indicate that our proposed clothing parsing method significantly improves processing time and training time over state-of-the-art methods, while ensuring comparable parsing accuracy and improving label recall rate.

With the exponential growth of web image data, image tagging is becoming crucial in many image based applications such as object recognition and content-based image retrieval. Despite the great progress achieved in automatic recognition technologies, none has yet provided a satisfactory solution to be widely useful in solving generic image recognition problems. So far, only manual tagging can provide reliable tagging results. However, such work is tedious, costly and workers have no motivation. In this paper, we propose an online image tagging system, EyeDentifyIt, driven by image-click-ads framework, which motivates crowdsourcing workers as well as general web users to tag images at high quality for low cost with low workload. A series of usability studies are presented to demonstrate how EyeDentifyIt provides improved user motivations and requires less workload, compared to state-of-the-art approaches.

Interactive visualization of medical data plays an increasingly vital role in clinical diagnosis. Cluttered views, lack of visual cues, etc. Are some of the problems that exist specially when projecting a 3D rendering of a Computed Tomography data on a 2D screen. In this paper we proposed two synthetic visual cues. They facilitate perception by removing visual clutter while maintaining context, directing viewers' attention to the object of interest, and consequently enhancing clinical diagnosis. Our proposed cues can be used both individually or combined with any other existing cues. Although our methods have been created with medical visualization in mind, it is possible to use them in other volume rendering applications as well.

With the exponential growth of web image data, image tagging is becoming crucial in many applications such as e-commerce. However, despite the great progress achieved in various image tagging technologies, none of them are able to incorporate browsing and discovery activities on web viewers in such a way that a user can easily query an image and ask the question "what is that in the image?". We have developed a comprehensive online image tagging system based on a Tagging-Tracking-Learning (TTL) framework to solve this problem. Tagging images using this system is able to turn common static web images into non-intrusive interactive images. The system tracks all browsing and interaction activities of users over time to filter out low quality tags and in turn helps the tagging process by alleviating manual operations. In this paper, we describe the implementation of the TTL framework and the novel algorithms developed. Usability studies of the system indicate that the TTL framework provides a better user experiences and simplifies the process of obtaining large tagged image collections over state-of-the-art approaches.

Tabletop systems provide a versatile space for collaboration, yet, in many cases, are limited by the inability to differentiate the interactions of simultaneous users. We present See Me, See You, a lightweight approach for discriminating user touches on a vision-based tabletop. We contribute a valuable characterization of finger orientation distributions of tabletop users. We exploit this biometric trait with a machine learning approach to allow the system to predict the correct position of users as they touch the surface. We achieve accuracies as high as 98% in simple situations and above 92% in more challenging conditions, such as two-handed tasks. We show high acceptance from users, who can self-correct prediction errors without significant costs. See Me, See You is a viable solution for providing simple yet effective support for multi-user application features on tabletops.

Colors images produced by current sensors are affected by many environmental factors resulting in the fact that even for the same illumination conditions, corresponding pixels in stereo pairs cannot be guaranteed to have the same color. In many cases, color based stereo matching is not a good choice to compute a good disparity maps. In this paper, we propose to solve this problem by using a relative gradient algorithm. Boundary and low texture problems are resolved by using a Gaussian weighting function and by limiting the search range. The experimental results show the proposed local method is effective and robust, and even outperforms some of the well-known global methods.

We present an experience-based approach to designing a collaborative interactive performance, humanaquarium. Our research explores public interaction with digital technology through the practice-based inquiry of an inter-disciplinary team of interaction designers and musicians. We present a method of designing experience from within, literally situating ourselves within the performance/use space and assuming the roles both of performers and of designers as we develop and refine the humanaquarium project over the course of a year's worth of public performances.

Real-time rendering of static volumetric data is generally known to be a memory and computationally intensive process. With the advance of graphic hardware, especially GPU, it is now possible to do this using desktop computers. However, with the evolution of real-time CT and MRI technologies, volumetric rendering is an even bigger challenge. The first one is how to reduce the data transmission between the main memory and the graphic memory. The second one is how to efficiently take advantage of the time redundancy which exists in time-varying volumetric data. We proposed an optimized compression scheme that explores the time redundancy as well as space redundancy of time-varying volumetric data. The compressed data is then transmitted to graphic memory and directly rendered by the GPU, reducing significantly the data transfer between main memory and graphic memory.

In this paper, we propose a novel framework to solve the image matting problem. We design a temporary image based on the estimated foreground and background colors for unknown pixels as well as an initial matte. The similarity of the temporary image and original image is modeled as an energy function in the Markov Random Field (MRF). The global optimized matte is obtained by minimizing the energy function. Therefore, image matting is converted to how to maximize the similarity of the original image and the temporary image. The experiments demonstrate that the proposed method could produce high quality mattes and it is also more effective compared to other top ranking methods.

People with a wide spectrum of disabilities, ranging from spinal injuries to autism, have benefited from equine assisted therapy (EAT). Using EAT, therapy patients have improved both physically and psychologically (e.g., demonstrating increased attention, motivation, and communication skills). There are still many open questions regarding this therapy and the reasons for its success. Many of these questions have remained unanswered due in large part to the uncontrolled nature of EAT. The Virtual Equine Assisted Therapy (VEAT) Project integrates a robotic platform with virtual reality technologies to provide a safe, controlled environment through which various aspects of EAT can be isolated and studied. The system incorporates realistic equine motions with visual, auditory, olfactory, and somatosensory stimuli to provide highly immersive experiences to patients.

Symmetric distributed source coding has made great progress in the past several years. However, few attempts have been tried to use the symmetric distributed source coding schemes in real data compression scenarios. In this paper, with the inspiration of previous symmetric distributed source coding schemes, we first put forward a novel syndrome-based symmetric distributed coding scheme which can achieve the whole Slepian-Wolf rate region. Then we propose a general architecture for symmetric distributed multiview video schemes. Our experiments show very encouraging results. In the high-rate case, our scheme can achieve great rate saving from separate H.264 coding scheme at the same video quality.

In the past decade, machine learning techniques have made great progress. Inspired by the recent advancement on semi-supervised learning techniques, we propose a novel learning-based multiview video compression framework. Our scheme can efficiently compress the multiview video represented by multiview-video-plus-depth (MVD) format.We model the multiview video compression problem as a semi-supervised learning problem and design sophisticated mechanisms to achieve high compression efficiency. Our approach is significantly different from the traditional hybrid coding scheme such as H.264-based multiview video coding methods. The preliminary results show promising compression performance.

We present an integrated interface that takes advantage of several VR technologies for the exploration of small artifacts in a Museum. This interface allows visitors to observe pieces in 3D from several viewpoints, touch them, feel their weight, and hear their sound at touch. From one hand, it will allow visitors to observe artifacts closer, in order to know more about selected pieces and to enhance their overall experience in a Museum. From the other hand, it leverages existing technologies in order to provide a multimodal interface, which can be easily replaced for others depending on costs or functionality. We describe some of the early results and experiences we provide in this setup.

The Gold Museum, in Bogota, Colombia, displays the largest collection of pre-Hispanic gold artifacts in the world and it has been renovated recently. With funds from the Colombian Government, we have created a multimodal experience that allows visitors to touch, hear, and see small artifacts. Here we present a description of this demo, its functionality, and technical requirements.

We present the iterative development and initial evaluation of a multi-modal platform for interacting with precious small artifacts from the Gold Museum in Bogota. By using a commercial haptic interface, loud speakers, and stereo displays, one can allow visitors to touch, hear, and observe in stereo those precious artifacts. We use this multi-modal interface in a novel way and in a novel context in order to provide virtual replicas that can be weighed, cleaned, and explored as if they were close to a visitor's hand. This platform is currently open to the public, and some of the lessons learned are reported in terms of usability in a real-world museum application.

Distributed source coding is a promising enabling technology for sensor network applications. Symmetric distributed source coding can achieve the entire Slepian-Wolf rate region and offer sensor network applications a wide range of options to compress and transmit data. In this paper, we design a new symmetric distributed coding scheme that realize the benefits of both simplified code construction and bit incorrespondence handling. In particular, we propose a general scheme to deal with the bit incorrespondence problem, which is not specifically addressed by previous approaches. With the ability to deal with the bit incorrespondence in the code structure, it is more likely for our approach to be used in real sensor network environments.

We propose a new solution to the problem of bi-layer video segmentation in terms of both, hardware design and algorithmic solution. At the data acquisition stage, we combine color video with infrared video, which is robust to illumination changes and provides an automatic initialization of the cue map for foreground-background segmentation. Two algorithms are presented to complete the segmentation, graph cut and contrast-preserving relaxation labeling. Both algorithms use color and edge information. Our experimental results show that the better performance of contrast-preserving relaxation labeling over graph cut, and the parallel characteristics make relaxation labeling superior to graph cut for implementation in GPU hardware.

Haptic interfaces are used increasingly in medical systems and related applications, but relatively little is known on the effectiveness of these interfaces. This paper reports a study on the perception of haptic force magnitude during hand movements. Discrimination thresholds were determined for a reference force of 1.5N in five different directions (0°, 45°, 90°, 135°, and 180°) with respect to the movement direction. We found that force discrimination thresholds detected were significantly higher during hand movement than those reported previously without hand movement, indicating that the perception of force magnitude is impaired by hand movement. The results also show there is no significant difference between the discrimination thresholds found for fast (28mm/s) and slow (14mm/s) hand movements. Finally, we found that the perception of force magnitude is impaired at a force direction of 45° with respect to the hand movement, indicating the existence of an oblique effect.

In this paper, we investigate two ways of employing infrared video with color video for automatic foreground-background video segmentation: foreground infrared (IR) illumination and background IR illumination. Foreground IR illumination gives an initial foreground template, which is combined with image segmentation to complete foreground segmentation. Two algorithms are explored, Graph Cut and Relaxation Labeling. The disadvantage of foreground IR illumination can be compensated by background illumination.

We propose an algorithm to produce a 3-D CAD model from a set of range data, based on non-uniform rational B-splines (NURBS) surface fitting technique. Our goal is to construct continuous geometric models, assuming that the topology of surface is unknown. In our approach, the triangulated surface is partitioned in quadrilateral patches, using Morse theory. The quadrilateral obtained mesh is regularized by means of the use of geodesic curves and B-splines to obtain a new adequate grid on which to draw NURBS surfaces. Such NURBS surfaces are optimized by means of evolutionary strategies. Further, the patches are smoothly joined guaranteeing continuity C 1 .

Multimedia content is becoming more widely used in applications resulting from the easy accessibility of high-speed networks in the public domain. An important component in multimedia content is 3D geometry, which in the past had low resolution due to acquisition, computational and network limitation, and was not able to approximate 3D surfaces realistically. Although processing speed and network capacity have been greatly increased in the last decade, the increase in demands for multimedia content surpass the increase in resources. Consequently, techniques for data simplification especially for 3D mesh data is inevitable in order to achieve shorter latency and satisfactory interactivity in applications. This paper presents a perceptual analysis to evaluate the visual quality associated with a change in level-of-detail. Our analysis is consistent to how the human visual system evaluates 3D objects in the real world and is based on the just-noticeable-difference methodology. Experimental results show that our approach presents an accurate estimation of visual quality and thus provides a systematic method to evaluate the performance of different simplification algorithms

Efficient online visualization of three-dimensional (3-D) mesh, mapped with photo realistic texture, is essential for a variety of applications such as museum exhibits and medical images. In these applications synthetic texture or color per vertex loses authenticity and resolution. An image-based view dependent approach requires too much overhead to generate a 360/spl deg/ display for online applications. We propose using a mesh simplification algorithm based on scale-space analysis of the feature point distribution, combined with an associated visual perception analysis of the surface texture, to address the needs of adaptive online transmission of high quality 3-D objects. The premise of the proposed textured mesh (TexMesh)simplification, taking the human visual system into consideration, is the following: given limited bandwidth, texture quality in low feature density surfaces can be reduced, without significantly affecting human perception. The advantage of allocating higher bandwidth, and thus higher quality, to dense feature density surfaces, is to improve the overall visual fidelity. Statistics on feature point distribution and their associated texture fragments are gathered during preprocessing. Online transmission is based on these statistics,which can be retrieved in constant time. Using an initial estimated bandwidth,a scaled mesh is first transmitted. Starting from a default texture quality,we apply an efficient Harmonic Time Compensation Algorithm based on the current bandwidth and a time limit, to adaptively adjust the texture quality of the next fragment to be transmitted. Properties of the algorithm are proved. Experimental results show the usefulness of our approach.

This paper presents a multiscale fluid model based on mesoscale dynamics and viscous fluid equations as a generic tool for digital marbling purposes. The model uses an averaging technique on the adaptation of a stochastic mesoscale model to obtain the effect of fluctuations at different levels. It allows various user controls to simulate complex flow behaviors as in traditional marbling techniques, as well as laminar and turbulent flows. Material transport is based on an improved advection solution to be able to match the highly detailed, sharp fluid interfaces in marbling patterns. In the transport model, two reaction models are introduced to create different effects and to simulate density fluctuations.

Multiview video compression is important to the image based 3D video applications. In this paper, we proposes a novel neighbor-based multiview video compression scheme. It is essentially a MPEG2-like block-based scheme. In particular, a method to decide the stream encoding order is presented. The resulting stream encoding order can better decorrelate spatial redundancies among multiple video streams than the center approach. Experimental results confirm the superiority of the proposed neighbor approach over the center approach and MPEG2 for multiview video compression.

Efficient online visualization of three-dimensional (3-D) textured models is essential for a variety of applications including not only games and e-commerce, but also heritage and medicine. To visualize 3-D objects online, it is necessary to quickly adapt both mesh and texture to the available computational or network resources. Earlier research showed that after reaching a minimum required mesh density, high-resolution texture has more impact on human perception than a denser mesh. Given limited bandwidth, an important issue is how to extract features that best represent the original object, and how to allocate resources between mesh and texture data to achieve optimal perceptual quality. In this paper, we propose a textured mesh (TexMesh) model, which applies scale-space analysis and perceptual evaluation to extract 3-D features for textured mesh simplification and transmission. Texture data is divided into fragments to facilitate quality and bandwidth adaptation. Texture quality assignment is based on feature point distribution. Online transmission is based on statistics gathered during preprocessing, which are stored in a priority queue and lookup tables. Quality of service requested by a client site can be met by applying an efficient adaptive algorithm to ensure optimal use of the specified time and available bandwidth, and at the same time preserving satisfactory quality. Our TexMesh framework integrates feature extraction, mesh simplification, texture reduction, bandwidth adaptation, and perceptual evaluation into a multiscale visualization framework.

The main idea behind tele-immersive environment is to create an immersive virtual environment that connect people across networks and enable them to interact not only with each other, but also with various other forms of shared digital data (video, 3D models, images, text, etc.). Tele-immersive environments may eventually replace current video and telephone conferencing, and enable for a better and more intuitive way to communicate between people and computer systems. To accomplish this, participants to a meeting has to be represented digitally with a high degree of accuracy in order to keep a sense of immersion. Tele-immersive environments should have the same "feel" as a real meeting. Interactions among people should be natural. In other to create such a system, we need to solve the key problem of how to create in real-time new views from a fixed network of cameras that will correspond to new viewpoints. We also need to do this for two virtual cameras corresponding to the inter-ocular distance of each participant. In this paper, we will describe a new binocular view interpolation method based on a re-projection technique using calibrated cameras. We will discuss the various aspects of this new algorithm and of the hardware systems necessary to perform these operations in real-time. We will also present early experimental results illustrating the various advantages of this algorithm.

This paper describes a reconfigurable VR environment and a markup language for creating experiments aimed at understanding human spatial navigation. It permits the creation of high-quality virtual environments and the recording of behavioral and brain activity measures while observers navigate these environments. The system is used in studies where the electroencephalographic activity is recorded while observers navigate virtual environments. The results of the study reported here confirmed previous finding that theta oscillations (electroencephalographic activity in the 4-8 Hz band) are linked to the difficulty of spatial navigation. Further, it showed that this activity is likely to occur at points where new rooms come into view, or after navigational mistakes have been realized and are being corrected. This indicates that theta oscillations in humans are related to the encoding and retrieval of spatial information.

Efficient online 3D visualization is essential for a variety of applications, including not only games and e-commerce, but also heritage and medicine. For efficient online visualization, it is necessary to adapt 3D models (both mesh and texture) quickly to the available computational or network resources. We propose using 3D model simplification based on a scale-space analysis of the surface curvature variations combined with an associated scale-space analysis of the surface texture to reduce the size of texture files, and facilitate distributed transmission. The premise of the proposed simplification is that minor variations in texture can be ignored in relatively smooth regions of a 3D surface, without significantly affecting human perception. Statistics of feature points and their associated texture fragments are gathered during preprocessing. On-line transmission and rendering for the next higher resolution scale is based on the statistics, which can be retrieved in constant time. Quality of service (QoS) can be provided based on the time limit, or the number of vertices, or faces, requested by the viewer. Experimental results showing the simplified models demonstrate the feasibility of our approach.

This paper describes the architecture of the ANIMUS framework. This framework facilitates the creation of synthetic characters that convey the illusion of being alive. The components of ANIMUS are inspired by observations made in biological organisms, and provide means for creating autonomous agents that mimic awareness of their environment, of other agents, and of human audience. They also show particular roles, personality, and emotions, active and reactive behavior, automatic reflexes, and selective attention; use temporal memory and learning capabilities to evolve in their dynamic virtual worlds, and express their thought and emotions with a flexible animation system while they interact with the user in immersive 3D environments. ANIMUS creatures follow artistic conceptual designs and constraints that determine the way they behave, react and interact with other creatures and the user, allowing the designer to create meaningful and interesting characters. The framework can be applied to complex immersive environments like CAVE systems or other interactive applications like video games and advanced man-machine interfaces, providing high level tools for creating a new generation of responsive believable agents.

We propose an automated inspection system of manufactured parts using a cloud of 3D measured points of a part provided by a range sensor, and its CAD model. Inspection consists in verifying the accuracy of a part related to a given set of tolerances. It is thus necessary that the 3D measurements be accurate. In the 3D capture of a part, several sources of error can alter the measured values. So, we have to find and model the most influential parameters affecting the accuracy of the range sensor in the digitalization process. This model is used to produce a sensing plan to acquire accurately the geometry of a part. By using the noise model, we introduce a dispersion value for each 3D point acquired. This value of dispersion is shown as a weight factor in the inspection results.

Virtual reality researchers and artists will discuss the use of virtual reality as a creative and presentation medium. Museums, galleries, libraries, theatres, research entres, theme parks make varied use of VR. VR is no longer gee whiz, strong content, effective management of technology and good presentation contexts are needed. How do different members of and groups within the public understand the v.r. experience? How can we afford VR for public spaces? What about networked experiences? What devices should we use, how can we create social space, personal experience? Join our panel of experts to explore the design, content and technology challenges of VR in public space.

Discusses the progress and the evolution of the development of range sensing techniques at the NRC laboratories. Essentially a 3D imaging project at the beginning, it has evolved to a new media project which requires the development of new tools for 3D modeling, editing, database searching and visualization. Generic applications related to documentation, inspection, target tracking and visual communication will be discussed.

A method of estimating range flow (space displacement vector field) on nonrigid as well as rigid objects from a sequence of range images is described. This method can directly estimate the deformable motion parameters by solving a system of linear equations that are obtained from substituting a linear transformation of nonrigid objects expressed by the Jacobian matrix into motion constraints based on an extension of the conventional scheme used in intensity image sequences. The range flow is directly computed by substituting these estimated motion parameters into the linear transformation. The algorithm is supported by experimental estimations of range flow on a sheet of paper, a piece of cloth, human skin, and a rubber balloon being inflated, using real range image sequences acquired by a video rate range camera.>

A method is described for estimating range flow (space displacement vector field) on a nonrigid object as well as a rigid object from a sequence of range images. The method can directly estimate the deformable motion parameters by solving a system of linear equations which are obtained by substituting a linear transformation of a nonrigid object expressed by a Jacobian matrix into motion constraints based on an extension of the conventional scheme used in intensity image sequences, and the range flow is obtained from the motion parameters and linear transformation. The algorithm is supported by experiments of motion estimation of a sheet of paper, a piece of cloth, the movement of human skin, and the motion of a rubber balloon being inflated using real range image sequences acquired by a video rate range finder. Furthermore, it is shown that by adding the intensity image sequence to the range information, one can prevent the equation system from falling in an ill-posed condition.>

A technique for the detection of depth and orientation discontinuities in range images is described. This method uses basic morphological operators to extract these discontinuities from a dense range map. The authors describe how to extract depth discontinuities using the morphological edge operator of J.S.J. Lee et al. (1987) and how to use a top-hat operator to localize the exact positions of these discontinuities. The detection of orientation discontinuities is then discussed in detail. The calculation of a good estimate of the surface normals using anisotropic filtering inhibited at depth discontinuities is demonstrated. Once these estimates of the normals are calculated, operations similar to those used in depth discontinuity detection are used to locate orientation discontinuities at abrupt variations of the surface normals. Experimental results demonstrate computation of the positions of these discontinuities at pixel precision as well as very efficient use of morphological operators to localize abrupt variations in a noisy image signal in less than 1 s using a Datacube image processor.>

An edge detection and classification scheme for range images which produces a multiscale representation in terms of well-localized depth and orientation edges is presented. The extraction is accomplished by detecting the presence of significant edges at a coarse scale and then determining their precise location by tracking them over decreasing scale. An adaptive multiscale thresholding is applied during this focusing process ro inhibit the attraction of insignificant details. Once focused, the edges are classified into the categories of true edge and diffuse edge by invoking classification rules derived from a mathematical analysis of edge displacement and branching over scale-space. Experimental results illustrate the robustness of the approach in the presence of noise and its performance with synthetic and real images of varying complexity. Comparisons with recently published techniques point out the improved performance of the approach, especially when the images contain substantially overlapping objects.>