2011
Permanent URI for this collection
tStudy of Augmented RealityMethods for Real Time Recognitionand Tracking of Untextured 3DModels in Monocular Images
[meta data] [files: ]
Alvarez Ponga, Hugo
Visual Semantic Analysis to Support Semi Automatic Modeling of Service Descriptions
[meta data] [files: ]
Bhatti, Nadeem
Pointing Facilitation Techniques for 3D Object Selection in Virtual Environments
[meta data] [files: ]
Argelaguet Sanz, Fernando
Symmetry in 3D Shapes - Analysis and Applications to Model Synthesis
[meta data] [files: ]
Bokeloh, Martin
Multiresolution Techniques for Real-Time Visualization of Urban Environments and Terrains
[meta data] [files: ]
Di Benedetto, Marco
Real-time GPU-accelerated Out-of-Core Rendering and Light-field Display Visualization for Improved Massive Volume Understanding
[meta data] [files: ]
Iglesias Guitian, Jose A.
Information-based Feature Enhancement in Scientific Visualization
[meta data] [files: ]
Haidacher, Martin
Browse
Recent Submissions
Item tStudy of Augmented RealityMethods for Real Time Recognitionand Tracking of Untextured 3DModels in Monocular Images(Alvarez Ponga, dec, 2011) Alvarez Ponga, HugoThe main challenge of an augmented reality system is to obtain perfectalignment between real and virtual objects in order to create the illusionthat both worlds coexist. To that end, the position and orientation of theobserver has to be determined in order to configure a virtual camera thatdisplays the virtual objects in their corresponding position. This problemis known as tracking, and although there are many alternatives to addressit by using different sensors, tracking based on optical sensors is the mostpopular solution. However, optical tracking is not a solved problem.This thesis presents a study of the existing optical tracking methodsand provides some improvements for some of them, particularly for thosethat are real time. More precisely, monocular optical marker tracking andmodel-based monocular optical markerless tracking are discussed in detail.The proposed improvements are focused on industrial environments, whichis a difficult challenge due to the lack of texture in these scenes.Monocular optical marker tracking methods do not support occlusions,so this thesis proposes two alternatives: (1) a new tracking method basedon temporal coherence, and (2) a new marker design. Both solutions arerobust against occlusions and do not require more environment adaptation.Similarly, the response of model-based monocular optical markerlesstracking methods is jeopardized in untextured scenes, so this thesis proposesa 3D object recognition method that uses geometric properties instead oftexture to initialize the tracking, as well as a markerless tracking methodthat uses multiple visual cues to update the tracking.Additionally, the details of the augmented reality system that has beendeveloped to help in disassembly operations are given throughout the thesis.This serves as a tool to validate the proposed methods and it also showstheir real world applicability.Item Visual Semantic Analysis to Support Semi Automatic Modeling of Service Descriptions(Bhatti, 2011-06-10) Bhatti, NadeemA new trend Web service ecosystems for Service-Oriented Architectures (SOAs) and Web services is emerging. Services can be offered and traded like products in these ecosystems. The explicit formalization of services' non-functional parameters, e.g. price plans and legal aspects, as Service Descriptions (SD) is one of the main challenges to establish such Web service ecosystems. The manual modeling of Service Descriptions (SDs) is a tedious and cumbersome task. In this thesis, we introduce the innovative approach Visual Semantic Analysis (VSA) to support semi-automatic modeling of service descriptions in Web service ecosystems. This approach combines the semantic analysis and interactive visualization techniques to support the analysis, modeling, and reanalysis of services in an iterative loop. For example, service providers can analyze first the price plans of the already existing services and extract semantic information from them (e.g. cheapest offers and functionalities). Then they can reuse the extracted semantics to model the price plans of their new services. Afterwards, they can reanalyze the new modeled price plans with the already existing services to check their market competitiveness in Web service ecosystems. The experts from different domains, e.g. service engineers, SD modeling experts, and price plan experts, were interviewed in a study to identify the requirements for the VSA approach. These requirements cover aspects related to the analysis of already exiting services and reuse of the analysis results to model new services. Based on the user requirements, we establish a generic process model for the Visual Semantic Analysis. It defines sub processes and transitions between them. Additionally, the technologies used and the data processed in these sub processes are also described. We present also the formal specification of this generic process model that serves as a basis for the conceptual framework of the VSA. A conceptual framework of the VSA elucidates structure and behavior of the Visual Semantic Analysis System. It specifies also system components of the VSA system and interaction between them. Additionally, we present the external interface of the VSA system for the communication with Web service ecosystems. Finally, we present the results of a user study conducted by means of the VSA system that is developed on the base of the VSA conceptual framework. The results of this user study show that the VSA system leads to strongly significant improvement of the time efficiency and offers better support for the analysis, modeling and reanalysis of service descriptions.Item Pointing Facilitation Techniques for 3D Object Selection in Virtual Environments(Argelaguet Sanz, 2011-07-04) Argelaguet Sanz, FernandoSelection is one of the fundamental tasks in virtual reality applications and the initial task for most common user's interactions in a virtual environment. In this thesis we analyze major factors influencing selection performance, and propose new techniques for facilitating selection in 3D space. Considering the frequency of selection tasks in a typical virtual reality workflow, improving selection tasks often results in significant gains in the overall user performance.A 3D selection task requires the user to gesture in 3D space, e.g. grabbing an object or pointing to something. The success or failure of the task depends mainly on the interaction technique, the dexterity of the user, and the spatial perception of the virtual environment. Since the dexterity of the user can be improved by training, we focus on how to take advantage of existing human control models to minimize the effort required to select an object, and how to enhance the user's spatial perception of the virtual environment to facilitate selection and referral tasks. We have proposed several selection techniques based on Fitts' Law and we have studied how visual feedback can be used to overcome spatial perception limitations in virtual environments. The techniques proposed are not only oriented to achieve performance gains as we also account for user's preferences. During the development of this thesis we have conducted a number of user studies, both to validate our theoretical analyses, and to compare the proposed selection techniques to existing ones.Although the major contributions of this thesis refer to the selection of 3D objects, we also provide new techniques for facilitating the interaction with 2D graphical user interfaces embedded in 3D space. We also explore selection tasks in collaborative virtual environments. In CVEs pointing tasks often change their purpose and turn into referring tasks. Referential awareness can be compromised in complex environments, because a user can point to a feature in the environment which might be occluded for the other users. We have analyzed how improvements on referential awareness increase the information exchange among users without violating social protocols in formal presentations.</p>Item Investigating Augmented Reality Visio-Haptic Techniques for Medical Training(Timothy R Coles, 2011-02-16) Coles, Timothy R<p>It is widely accepted that a reform in medical teaching must be made to meet today s high volume training requirements. Receiving pre-training in a core set of surgical skills and procedures before novice practitioners are exposed to the traditional apprenticeship training model where an experienced practitioner must always be present, can reduce both skill acquisition time and the risks patients are exposed to due to surgeon inexperience. Virtual simulation offers a potential method of providing this training and a subset of current medical training simulations integrate haptics and visual feedback to enhance procedural learning.</p><p>The role of virtual medical training applications, in particular where haptics (force and tactile feedback) can be used to assist a trainee to learn and practice a task, is investigated in this thesis. A review of the current state-of-the-art summarises considerations that must be made during the deployment of haptics and visual technologies in medical training, including an assessment of the available force/torque, tactile and visual hardware solutions in addition to the haptics related software. An in-depth analysis of medical training simulations that include haptic feedback is then provided after which the future directions and current technological limitations in the field are discussed.</p><p>The potential benefits of developing and using a new Augmented Reality (AR) visio-haptic medical training environment is subsequently explored, and an exemplar application called PalpSim has been produced to train femoral palpation and needle insertion, the opening steps of many Interventional Radiology (IR) procedures. This has been performed in collaboration with IR experts. PalpSim s AR environment permits a trainee to realistically interact with a computer generated patient using their own hands as if the patient existed in the real world. During a simulation, the trainee can feel haptic feedback developed from in vivo measured force data whilst palpating deformable tissue and inserting a virtual needle shaft into a simulated femoral artery, at which point virtual blood flow from the real needle hub will be seen. The PalpSim environment has undergone face and content validation at the Royal Liverpool University Hospital and received positive feedback.</p><p>An important requirement identified was for a haptics device combining force and tactile feedback to closely simulate the haptic cues felt during femoral palpation. Two cost effective force feedback devices have therefore been modified to provide the degrees of force feedback needed to closely recreate the forces of a palpation procedure and are combined with a custom built hydraulic tactile interface to provide pulse-like tactile cues. A needle interface based on a modified PHANTOM Omni also allows the user to grasp and see a real interventional radiology needle hub whilst feeling simulated insertion forces.</p><p>PalpSim is the first example of a visio-haptic medical training environment based on chroma-key augmented reality technology. It is expected that many other medical training solutions will adopt this approach in the future.</p>Item Symmetry in 3D Shapes - Analysis and Applications to Model Synthesis(Bokeloh, 2011-11-29) Bokeloh, MartinSymmetry is an essential property of a shapes' appearance and presents a source of information for structure-aware deformation and model synthesis. This thesis proposes feature-based methods to detect symmetry and regularity in 3D shapes and demonstrates the utilization of symmetry information for content generation. First, we will introduce two novel feature detection techniques that extract salient keypoints and feature lines for a 3D shape respectively. Further, we will propose a randomized, feature-based approach to detect symmetries and decompose the shape into recurring building blocks. Then, we will present the concept of docking sites that allows us to derive a set of shape operations from an exemplar and will produce similar shapes. This is a key insight of this thesis and opens up a new perspective on inverse procedural modeling. Finally, we will present an interactive, structure-aware deformation technique based entirely on regular patterns.Item Visually Pleasing Real-time Global Illumination Rendering for Fully-dynamic Scenes(Dong, 2011-03-03) Dong, ZhaoGlobal illumination (GI) rendering plays a crucial role in the photo-realistic rendering of virtual scenes. With the rapid development of graphics hardware, GI has become increasingly attractive even for real-time applications nowadays. However, the computation of physically-correct global illumination is time-consuming and cannot achieve real-time, or even interactive performance. Although the real-time GI is possible using a solution based on precomputation, such a solution cannot deal with fully-dynamic scenes. This dissertation focuses on solving these problems by introducing visually pleasing real-time global illumination rendering for fully-dynamic scenes. To this end, we develop a set of novel algorithms and techniques for rendering global illumination effects using the graphics hardware. All these algorithms not only result in real-time or interactive performance, but also generate comparable quality to the previous works in off-line rendering. First, we present a novel implicit visibility technique to circumvent expensive visibility queries in hierarchical radiosity by evaluating the visibility implicitly. Thereafter, we focus on rendering visually plausible soft shadows, which is the most important GI effect caused by the visibility determination. Based on the pre-fltering shadow mapping theory, we successively propose two real-time soft shadow mapping methods:Item Multiresolution Techniques for Real-Time Visualization of Urban Environments and Terrains(Di Benedetto, 2011-06-21) Di Benedetto, MarcoIn recent times we are witnessing a steep increase in the availability of data comingfrom real-life environments. Nowadays, virtually everyone connected to the Internetmay have instant access to a tremendous amount of data coming from satelliteelevation maps, airborne time-of-flight scanners and digital cameras, street-levelphotographs and even cadastral maps. As for other, more traditional types of mediasuch as pictures and videos, users of digital exploration softwares expect commodityhardware to exhibit good performance for interactive purposes, regardless of thedataset size.In this thesis we propose novel solutions to the problem of rendering large terrainand urban models on commodity platforms, both for local and remote exploration.Our solutions build on the concept of multiresolution representation, where alternativerepresentations of the same data with different accuracy are used to selectivelydistribute the computational power, and consequently the visual accuracy, where itis more needed on the base of the user's point of view.In particular, we will introduce an efficient multiresolution data compression techniquefor planar and spherical surfaces applied to terrain datasets which is able tohandle huge amount of information at a planetary scale. We will also describe a noveldata structure for compact storage and rendering of urban entities such as buildingsto allow real-time exploration of cityscapes from a remote online repository.Moreover, we will show how recent technologies can be exploited to transparentlyintegrate virtual exploration and general computer graphics techniques with webapplications.Item Error-Concealed Image-based Rendering(Eisemann, 2011-07-06) Eisemann, MartinCreating photo-realistic images has been one of the major goals in computer graphics since its early days. Instead of modeling the complexity of nature with standard modeling tools, imagebased approaches aim at exploiting real-world footage directly, as they are photo-realistic by definition. A drawback of these approaches has always been that the composition or combination of different sources is a non-trivial task, often resulting in annoying visible artifacts. In this thesis we focus on different techniques to diminish visible artifacts when combining multiple images in a common image domain. The results are either novel images, when dealing with the composition task of multiple images, or novel video sequences rendered in real-time, when dealing with video footage from multiple cameras.Item Visualization and Graphical Processing of Volume Data(Gavrilescu, 2011-10-17) Gavrilescu, MariusThe extraction and visualization of information from volume data constitute important research avenues in computer graphics and imaging. The rapid development of GPUs with increasing computational power has made it possible to generate intuitive, three-dimensional representations of data sets, which can be tweaked and manipulated interactively. This thesis presents various techniques developed within the field of volume graphics. These have wide applicability in the generation of meaningful images from mainly CT and MRI data sets. The work addresses multiple aspects of volume visualization and rendering, such as the representation, classification and in-depth graphical analysis of the information contained within volume data. Initially, we present generic information on the nature of volume data, the mathematical and physical models behind volume rendering, as well as the rendering algorithms used within our prototyping framework for the rendering of images. Subsequently, we address the problem of volume classification, where we explore the use of various types of transfer functions. These operate on voxel properties such as the gradient, curvature or visibility, allowing for the isolation of increasingly complex and problematic features. We provide alternative, more computationally-efficient ways of approximating some of these properties and show how they can be used for classification purposes. We also provide an effective way of specifying multidimensional transfer functions from 1D components, thus increasing the flexibility and expanding the potential of the classification process.Another part of the thesis deals with cardiac MRI data. Specifically, we develop a tool for the visual inspection of parameters which influence the status and functionality of the left ventricle. The considered parameters are the thickness and thickening of the myocardial wall, the moment of maximum thickness and the average speed of the wall during a cardiac cycle. Starting from segmentation contours which outline the epicardium and endocardium, we construct surfaces and use these to visualize the distribution of parameter values using color coding. The technique allows for information from multiple slices, over multiple phases and stress levels to be represented on a single 3D geometry, therefore facilitating the analysis of multidimensional data sets comprising a large number of slices. The values of the cardiac parameters are depicted in an intuitive manner, making them easily accessible to both medical staff and patients with no medical training. In the last part of the thesis we develop a method for the analysis of parameters involved in the volume rendering pipeline. The technique involves sampling the parameters across their domains, rendering images for each sample, and computing the differences among these images. The resulting values characterize the behavior and stability of the parameters across their domains. These values are further used to augment various user interfaces, such as sliders or transfer function specification widgets. The newly-modified interfaces use color coding, graphs, arrows and other info-vis techniques to show the potential changes induced by the parameters in images resulting from volume rendering, thus allowing users to make better-informed decisions when adjusting parameter values.Item Real-time GPU-accelerated Out-of-Core Rendering and Light-field Display Visualization for Improved Massive Volume Understanding(Iglesias Guitian, Jose A., 2011-03-07) Iglesias Guitian, Jose A.Nowadays huge digital models are becoming increasingly available for a number of different applications ranging from CAD, industrial design to medicine and natural sciences. Particularly, in the field of medicine, data acquisition devices such as MRI or CT scanners routinely produce huge volumetric datasets. Currently, these datasets can easily reach dimensions of 1024^3 voxels and datasets larger than that are not uncommon.This thesis focuses on efficient methods for the interactive exploration of such large volumes using direct volume visualization techniques on commodity platforms. To reach this goal specialized multi-resolution structures and algorithms, which are able to directly render volumes of potentially unlimited size are introduced. The developed techniques are output sensitive and their rendering costs depend only on the complexity of the generated images and not on the complexity of the input datasets. The advanced characteristics of modern GPGPU architectures are exploited and combined with an out-of-core framework in order to provide a more flexible, scalable and efficient implementation of these algorithms and data structures on single GPUs and GPU clusters.To improve visual perception and understanding, the use of novel 3D display technology based on a light-field approach is introduced. This kind of device allows multiple naked-eye users to perceive virtual objects floating inside the display workspace, exploiting the stereo and horizontal parallax. A set of specialized and interactive illustrative techniques capable of providing different contextual information in different areas of the display, as well as an out-of-core CUDA based ray-casting engine with a number of improvements over current GPU volume ray-casters are both reported. The possibilities of the system are demonstrated by the multi-user interactive exploration of 64-GVoxel datasets on a 35-MPixel light-field display driven by a cluster of PCs.Item Information-based Feature Enhancement in Scientific Visualization(Haidacher, 2011) Haidacher, MartinScientific visualization is a research area which gives insight into volumetric data acquired through measurement or simulation. The visualization allows a faster and more intuitive exploration of the data. Due to the rapid development in hardware for the measurement and simulation of scientific data, the size and complexity of data is constantly increasing. This has the benefit that it is possible to get a more accurate insight into the measured or simulated phenomena. A drawback of the increasing data size and complexity is the problem of generating an expressive representation of the data. Since only certain parts of the data are necessary to make a decision, it is possible to mask parts of the data along the visualization pipeline to enhance only those parts which are important in the visualization. For the masking various properties are extracted from the data which are used to classify a part as important or not. In general a transfer function is used for this classification process which has to be designed by the user. In this thesis three novel approaches are presented which use methods from information theory and statistics to enhance features from the data in the classification process that are important for a certain task. With the tools of information theory and statistics it is possible to extract properties from the data which are able to classify different materials or tissues in the data better than comparable other approaches. One approach adaptively extracts statistical properties, i.e. the mean value and the standard deviation, of the data values in the local neighborhood of each point in the data set. With these statistical properties it is possible to better distinguish between different materials in a data set even though the data is very noisy. The other two approaches in this thesis employ methods from information theory to extract features from multimodal data sets. Thus it is possible to enhance features of the data which are either very similar or very dissimilar in both modalities. Through information theory the variations in the value ranges of both modalities do not influence the classification of these features. All three approaches define novel transfer-function spaces which simplify the design process of a transfer function for the user. Different features of the data, such as different materials, can be clearly depicted in these spaces. Therefore, it is easier for a user to design a transfer function which enhances the features of importance for a certain task. For each of the new approaches results and comparisons to other existing techniques are shown to highlight the usefulness of the proposed methods. Through the described research it is shown that information theory and statistics are tools which are able to extract expressive properties from the data. In the introduction a broad overview over scientific visualization and the visualization pipeline is given. The classification process is described in more detail. Since information theory and statistics play an important role for all three approaches, a brief introduction to these concepts is given as well.Item New Models for High-Quality Surface Reconstruction and Rendering(Kalbe, 2011-03-17) Kalbe, ThomasThe efficient reconstruction and artifact-free visualization of surfaces from measured real-world data is an important issue in various applications, such as medical and scientific visualization, quality control, and the media-related industry. The main contribution of this thesis is the development of the first efficient GPU-based reconstruction and visualization methods using trivariate splines, i.e., splines defined on tetrahedral partitions. Our methods show that these models are very well-suited for real-time reconstruction and high-quality visualizations of surfaces from volume data. We create a new quasi-interpolating operator which for the first time solves the problem of finding a globally C1-smooth quadratic spline approximating data and where no tetrahedra need to be further subdivided. In addition, we devise a new projection method for point sets arising from a sufficiently dense sampling of objects. Compared with existing approaches, high-quality surface triangulations can be generated with guaranteed numerical stability. Keywords. Piecewise polynomials; trivariate splines; quasi-interpolation; volume data; GPU ray casting; surface reconstruction; point set surfacesItem Interactive Visual Analysis of Multi-faceted Scientific Data(Kehrer, 2011-03-01) Kehrer, Johannes<p>Visualization plays an important role in exploring, analyzing and presentinglarge and heterogeneous scientific data that arise in many disciplines ofmedicine, research, engineering, and others. We can see that model and data scenariosare becoming increasingly <i>multi-faceted</i>: data are often multi-variate andtime-dependent, they stem from different data sources (multi-modal data), frommultiple simulation runs (multi-run data), or from multi-physics simulations ofinteracting phenomena that consist of coupled simulation models (multi-modeldata). The different data characteristics result in special challenges for visualizationresearch and interactive visual analysis. The data are usually large andcome on various types of grids with different resolution that need to be fused inthe visual analysis.</p><p>This thesis deals with different aspects of the interactive visual analysis ofmulti-faceted scientific data. The main contributions of this thesis are: 1) anumber of novel approaches and strategies for the interactive visual analysis ofmulti-run data; 2) a concept that enables the feature-based visual analysis acrossan interface between interrelated parts of heterogeneous scientific data (includingdata from multi-run and multi-physics simulations); 3) a model for visual analysisthat is based on the computation of traditional and robust estimates of statisticalmoments from higher-dimensional multi-run data; 4) procedures for visualexploration of time-dependent climate data that support the rapid generationof promising hypotheses, which are subsequently evaluated with statistics; and5) structured design guidelines for glyph-based 3D visualization of multi-variatedata together with a novel glyph. All these approaches are incorporated in a singleframework for interactive visual analysis that uses powerful concepts such ascoordinated multiple views, feature specification via brushing, and focus+contextvisualization. Especially the data derivation mechanism of the framework hasproven to be very useful for analyzing different aspects of the data at differentstages of the visual analysis. The proposed concepts and methods are demonstratedin a number of case studies that are based on multi-run climate data anddata from a multi-physics simulation.</p>Item Visual Computing in Virtual Environments(Lancelle, 2011) Lancelle, MarcelThis thesis covers research on new and alternative ways of interactionwith computers. Virtual Reality and multi touch setupsare discussed with a focus on three dimensional renderingand photographic applications in the field of Computer Graphics.Virtual Reality (VR) andVirtual Environments (VE)were once thoughtto be the future interface to computers. However, a lot of problemsprevent an everyday use. This work shows solutions to some of theproblems and discusses remaining issues.Hardware for Virtual Reality is diverse and many new devices arestill being developed. An overview on historic and current devicesand VE setups is given and our setups are described. The DAVE, animmersive projection room, and the HEyeWall Graz, a large highresolution display with multi touch input are presented. Availableprocessing power and in some parts rapidly decreasing prices lead toa continuous change of the best choice of hardware. A major influenceof this choice is the application. VR and multi touch setups oftenrequire sensing or tracking the user, optical tracking being a commonchoice. Hardware and software of an optical 3D marker tracking andan optical multi touch system are explained.The Davelib, a software framework for rendering 3D models in VirtualEnvironments is presented. It allows to easily port existing 3D applicationsto immersive setups with stereoscopic rendering and headtracking. Display calibration and rendering issues that are special toVR setups are explained. User interfaces for navigation and manipulationare described, focusing on interaction techniques for the DAVEand for multi touch screens. Intuitive methods are shown that areeasy to learn and use, even for computer illiterates. Exemplary applicationsdemonstrate the potential of immersive and non-immersivesetups, showing which applications can most benefit from Virtual Environments.Also, some image processing applications in the area ofcomputational photography are explained, that help to better depictthe captured scene.Item Perception Based Image Editing(Lopez-Moreno, 2011-04-08) Lopez-Moreno, JorgeImage editing and post-processing techniques have matured over the years, making it difficult (vergingon impossible) to assess whether an image has been digitally enhanced or modiffied somehow. However,complex manipulations are still a time consuming process which relies on skilled user input, oftenrequiring painstakingly painting over pixels. In this thesis we present our work on advanced image editing techniques, extending current toolsby leveraging the limitations of the human visual system in order to extract additional dimensions(like depth or texture) from a single two-dimensional image. Working in perceptual space, the validityof our results is assessed by psychophysical methodologies.</p><p>image editing, perception, light source estimation, surface reconstruction, intrinsic imagesItem Optimization of Inverse Reflector Design(Mas, 2011-03-30) Mas, Albert<p>This thesis presents new methods for the inverse reflector design problem.We have focused on three main topics that are deeply related with theproblem: the use of real and complex light sources, the definition of a fast lighting simulation algorithm to compute the reflector lighting in a fast way, and the definition of an optimization algorithm to more efficiently find the desired reflector.</p><p>To achieve accuracy and realism in the lighting simulations, we have used near-field light sources.We present a method for compressing measured datasets of the near-field emission of physical light sources.We create a mesh on the bounding surface of the light source that stores illumination information.The mesh is augmented with information about directional distribution and energy density.We have developed a new approach to smoothly generate random samples on the illumination distribution represented by the mesh, and to efficiently handle importance sampling of points and directions.We show that our representation can compress a 10 million particle rayset into a mesh of a few hundred triangles.We also show that the error of this representation is low, even for very close objects.</p><p>Then, we have proposed a fast method to obtain the outgoing light distribution of a parameterized reflector, and then compare it with the desired illumination.The new method works completely in the GPU.We trace millions of rays using a hierarchical height-field representation of the reflector, and multiple reflections are taken into account.We show that our method can calculate a reflector lighting at least one order of magnitude faster than previous methods, even with millions of rays, complex geometries and light sources.</p><p>Finally, a new global optimization algorithm has been specifically tailored to minimize the function that calculates the difference between the reflector lighting and the desired one.The optimization is an iterative process where each step evaluates the difference between a reflector illumination and the desired one.We have proposed a tree-based stochastic method that drives the optimization process, using heuristic rules to reach a minimum below a threshold that satisfies the user-provided requirements.We show that our method reaches a solution in less steps than most other classic optimization methods, also avoiding many local minima.</p>Item Multi-Field Visualization(Obermaier, 2011-02-09) Obermaier, HaraldModern science utilizes advanced measurement and simulation techniques to analyzephenomena from fields such as medicine, physics, or mechanics. The dataproduced by application of these techniques takes the form of multi-dimensionalfunctions or fields, which have to be processed in order to provide meaningfulparts of the data to domain experts. Definition and implementation of such processingtechniques with the goal to produce visual representations of portions ofthe data are topic of research in scientific visualization or multi-field visualizationin the case of multiple fields.In this thesis, we contribute novel feature extraction and visualization techniquesthat are able to convey data from multiple fields created by scientificsimulations or measurements. Furthermore, our scalar-, vector-, and tensor fieldprocessing techniques contribute to scattered field processing in general and introducenovel ways of analyzing and processing tensorial quantities such as strainand displacement in flow fields, providing insights into field topology.We introduce novel mesh-free extraction techniques for visualization of complexvaluedscalar fields in acoustics that aid in understanding wave topology in lowfrequency sound simulations. The resulting structures represent regions withlocally minimal sound amplitude and convey wave node evolution and soundcancellation in time-varying sound pressure fields, which is considered an importantfeature in acoustics design.Furthermore, methods for flow field feature extraction are presented that facilitateanalysis of velocity and strain field properties by visualizing deformation ofinfinitesimal Lagrangian particles and macroscopic deformation of surfaces andvolumes in flow. The resulting adaptive manifolds are used to perform flow fieldsegmentation which supports multi-field visualization by selective visualizationof scalar flow quantities.The effects of continuum displacement in scattered moment tensor fields can bestudied by a novel method for multi-field visualization presented in this thesis.The visualization method demonstrates the benefit of clustering and separateviews for the visualization of multiple fields.Item Large Data Scalability in Interactive Visual Analysis(Piringer, 2011) Piringer, HaraldIn many areas of science and industry, the amount of data is growing fast and often already exceeds the ability to evaluate it. On the other hand, the unprecedented amount of available data bears an enormous potential for supporting decision-making. Turning data into comprehensible knowledge is thus a key challenge of the 21st century. The power of the human visual system makes visualization an appropriate method to comprehend large data. In particular interactive visualization enables a discourse between the human brain and the data that can transform a cognitive problem to a perceptual one. However, the visual analysis of large and complex datasets involves both visual and computational challenges. Visual limits involve perceptual and cognitive limitations of the user and restrictions of the display devices while computational limits are related to the computational complexity of the involved algorithms. The goal of this thesis is to advance the state of the art in visual analysis with respect to the scalability to large datasets. Due to the multifaceted nature of scalability, the contributions span a broad range to enhance computational scalability, to improve the visual scalability of selected visualization approaches, and to support an analysis of high-dimensional data. Concerning computational scalability, this thesis describes a generic architecture to facilitate the development of highly interactive visual analysis tools using multi-threading. The architecture builds on the separation of the main application thread and dedicated visualization threads, which can be cancelled early due to user interaction. A quantitative evaluation shows fast visual feedback during continuous interaction even for millions of entries. Two variants of scatterplots address the visual scalability of different types of data and tasks. For continuous data, a combination of 2D and 3D scatterplots intends to combine the advantages of 2D interaction and 3D visualization. Several extensions improve the depth perception in 3D and address the problem of unrecognizable point densities in both 2D and 3D. For partly categorical data, the thesis contributes Hierarchical Difference Scatterplots to relate multiple hierarchy levels and to explicitly visualize differences between them in the context of the absolute position of pivoted values. While comparisons in Hierarchical Difference Scatterplots are only qualitative, this thesis also contributes an approach for quantifying subsets of the data by means of statistical moments for a potentially large number of dimensions. This approach has proven useful as an initial overview as well as for a quantitative comparison of local features like clusters. As an important application of visual analysis, the validation of regression models also involves the scalability to multi-dimensional data. This thesis describes a design study of an approach called HyperMoVal for this task. The key idea is to visually relate n-dimensional scalar functions to known validation data within a combined visualization. The integration with other multivariate views is a step towards a user-centric workflow for model building. Being the result of collaboration with experts in engine design, HyperMoVal demonstrates how visual analysis is suitable to significantly improve real-world tasks. Positive user feedback suggests a high impact of the contributions of this thesis also outside the visualization research community. Moreover, most contributions of this thesis have been combined in a commercially distributed software framework for engineering applications that will hopefully raise the awareness and promote the use of visual analysis in multiple application domains.Item Image processing techniques and segmentation evaluation(Smochina, 2011-10-17) Smochina, CristianThis thesis presents contributions in the field of microscopic image analysis, in particularthe automatic segmentation of fluorescent images of cell nuclei and colon crypts. Theevaluation methodology of the segmentation results is detailed and a new evaluation criterionis presented.The proposed discrepancy method is based on the comparison: machine segmentation vs.ground-truth segmentation. This error measure eliminates the inconveniences that appear inthe case of concave objects and allows easy control of the method sensibility regarding theobjects shape similarity according to the field in which it is used.An analysis of the most used image processing methods in microscopic imagesegmentation is presented by considered both the pathological fields: cytology and histology.Segmentation methods are also proposed for both fields: segmentation of the nuclei (used incytometry) and crypts segmentation (used in hystometry).Since the critical problem in microscopic images from tissues with colon carcinoma isthe touching nuclei, three techniques are proposed to find the boundaries oftouching/clustered nuclei. Since all methods need accurate background delineation, twoapproaches are proposed for this matter.The segmentation problem of specific chained configurations is solved using the pointswith high concavity and a set of templates and rules to validate and to pair these points. Theclustered/touching cell nuclei within complex structures are separated using the shape of thesection profile or a cross-correlation with a specific template of the separation areas.Regarding the histological structures, two automatic segmentation techniques robustlyidentify the epithelial layer/crypts. Both proposed methods use hierarchical approaches likemorphological hierarchy or anisotropic diffusion pyramid. A useful study of the samplingstep and a comparison between the hierarchy (without sampling) and the pyramid (withsampling) is presented. The significant implication of these techniques consists of the coarseto-fine approach. First the high level information is preferred against the local one to allow aneasy detection of the positions for the interest objects. Next, a more detailed analysis of thehierarchical representations is performed in order to obtain an accurate segmentation.The evaluation has been done by comparison against ground-truth segmentations or byvisual inspecting by a human expert. The results confirmed that the proposed methods couldefficiently solve the segmentation problems of microscopic images.Item Deformable Shape Matching(Tevs, Art, 2011-12-21) Tevs, ArtDeformable shape matching has become an important building block inacademia as well as in industry.Given two three dimensional shapes A and B the deformation function faligning A with B has to be found.The function is discretized by a set of corresponding point pairs.Unfortunately, the computation cost of a brute-force search ofcorrespondences is exponential. Additionally, to be of any practicaluse the algorithm has to be able to deal with data coming directlyfrom 3D scanner devices which suffers from acquisition problems likenoise, holes as well as missing any information about topology.This dissertation presents novel solutions for solving shape matching:First, an algorithm estimating correspondences using a randomizedsearch strategy is shown. Additionally, a planning step dramaticallyreducing the matching costs is incorporated. Using ideas of these bothcontributions, a method for matching multiple shapes at once is shown.The method facilitates the reconstruction of shape and motion fromnoisy data acquired with dynamic 3D scanners. Considering shapematching from another perspective a solution is shown using MarkovRandom Fields (MRF). Formulated as MRF, partial as well as fullmatches of a shape can be found. Here, belief propagation is utilizedfor inference computation in the MRF. Finally, an approachsignificantly reducing the space-time complexity of belief propagationfor a wide spectrum of computer vision tasks is presented.Item Reconstructive Geometry(Ullrich, 2011) Ullrich, TorstenThe thesis “Reconstructive Geometry” by TORSTEN ULLRICH presents a new collision detection algorithm, a novel approach to generative modeling, and an innovative shape recognition technique. All these contributions are centered around the questions “howto combine acquisition data with generative model descriptions” and “how to perform this combination efficiently”. Acquisition data – such as point clouds and triangle meshes – are created e.g. by a 3D scanner or a photogrammetric process. They can describe a shape’s geometry very well, but do not contain any semantic information. With generative descriptions it’s the other way round: a procedure describes a rather ideal object and its construction process. This thesis builds a bridge between both types of geometry descriptions and combines them to a semantic unit. An innovative shape recognition technique, presented in this thesis, determines whether a digitized real-world object might have been created by a given generative description, and if so, it identifies the high-level parameters that have been passed to the generative script. Such a generative script is a simple JavaScript function. Using the generative modeling compiler “Euclides” the function can be understood in a mathematical sense; i.e. it can be differentiated with respect to its input parameters, it can be embedded into an objective function, and it can be optimized using standard numerical analysis. This approach offers a wide range of applications for generative modeling techniques; parameters do not have to be set manually – they can be set automatically according to a reasonable objective function. In case of shape recognition, the objective function is distance-based and measures the similarity of two objects. The techniques that are used to efficiently perform this task (space partitioning, hierarchical structures, etc.) are the same in collision detection where the question, whether two objects have distance zero, is answered. To sum up, distance functions and distance calculations are a main part of this thesis along with their application in geometric object descriptions, semantic enrichment, numerical analysis and many more.Item Visual Steering to Support Decision Making in Visdom(Waser, 2011-06-15) Waser, JürgenComputer simulation has become an ubiquitous tool to investigate the nature of systems. When steering a simulation, users modify parameters to study their impact on the simulation outcome. The ability to test alternative options provides the basis for interactive decision making. Increasingly complex simulations are characterized by an intricate interplay of many heterogeneous input and output parameters. A steering concept that combines simulation and visualization within a single, comprehensive system is largely missing. This thesis targets the basic components of a novel integrated steering system called Visdom to support the user in the decision making process. The proposed techniques enable users to examine alternative scenarios without the need for special simulation expertise. To accomplish this, we propose World Lines as a management strategy for multiple, related simulation runs. In a dedicated view, users create and navigate through many simulation runs. New decisions are included through the concept of branching. To account for uncertain knowledge about the input parameters, we provide the ability to cover full parameter distributions. Via multiple cursors, users navigate a system of multiple linked views through time and alternative scenarios. In this way, the system supports comparative visual analysis of many simulation runs. Since the steering process generates a huge amount of information, we employ the machine to support the user in the search for explanations inside the computed data. Visdom is built on top of a data-flow network to provide a high level of modularity. A decoupled meta-flow is in charge of transmitting parameter changes from World Lines to the affected dataflow nodes. To direct the user attention to the most relevant parts, we provide dynamic visualization inside the flow diagram. The usefulness of the presented approach is substantiated through case studies in the field of flood management. The Visdom application enables the design of a breach closure by dropping sandbags in a virtual environment.Item Multimodal Training of Maintenance andAssembly Skills Based on Augmented Reality(Webel, 2011-12-07) Webel, SabineThe training of technicians in the acquisition of new maintenance and assembly tasks is an importantfactor in industry. As the complexity of these tasks can be enormous, the training of technicians toacquire the necessary skills to perform them efficiently is a challenging point. However, traditionaltraining programs are usually highly theoretical and it is difficult for the trainees to transfer the acquiredtheoretical knowledge about the task to the real task conditions, or rather, to the physical performanceof the task. In addition, traditional training programs are often expensive in terms of effort and cost.Previous research has shown that Augmented Reality is a powerful technology to support training inthe particular context of industrial service procedures, since instructions on how to perform the servicetasks can be directly linked to the machine parts to be processed. Various approaches exist, in whichthe trainee is guided step-by-step through the maintenance task, but these systems act more as guidingsystems than as training systems and focus only on the trainees sensorimotor capabilities. Due to theincreasing complexity of maintenance tasks, it is not sufficient to train the technicians execution ofthese tasks, but rather to train the underlying skills sensorimotor and cognitive that are necessaryfor an efficient acquisition and performance of new maintenance operations.All these facts lead to the need for efficient training systems for the training of maintenance andassembly skills which accelerate the technicians learning and acquisition of new maintenance procedures.Furthermore, these systems should improve the adjustment of the training process to newtraining scenarios and enable the reuse of existing training material that has proven its worth. In thisthesis a novel concept and platform for multimodal Augmented Reality-based training of maintenanceand assembly skills is presented. This concept includes the identification of necessary sub-skills, thetraining of the involved skills, and the design of a training program for the training of maintenance andassembly skills. Since procedural skills are considered as the most important skills for maintenanceand assembly operations, they are discussed in detail, as well as appropriate methods for improvingthem. We further show that the application of Augmented Reality technologies and the provisionof multimodal feedback and vibrotactile feedback in particular have a great potential to enhanceskill training in general. As a result, training strategies and specific accelerators for the training ofmaintenance and assembly skills in general and procedural skills in particular are elaborated. Here,accelerators are concrete methods used to implement the pursued training strategies.Furthermore, a novel concept for displaying location-dependent information in Augmented Realityenvironments is introduced, which can compensate tracking imprecisions. In this concept, the pointercontentmetaphor of annotating documents is transferred to Augmented Reality environments. As aresult, Adaptive Visual Aids are defined which consist of a tracking-dependent pointer object and atracking-independent content object, both providing an adaptable level and type of information. Thus, the guidance level of Augmented Reality overlays in AR-based training applications can be easilycontrolled. Adaptive Visual Aids can be used to substitute traditional Augmented Reality overlays (i.e.overlays in form of 3D animations), which highly suffer from tracking inaccuracies.The design of the multimodal AR-based training platform proposed in this thesis is not specific forthe training of maintenance and assembly skills, but is a general design approach for multimodal trainingplatforms. We further present an implementation of this platform based on the X3D ISO standardwhich provides features that are useful for the development of Augmented Reality environments. Thisstandard-based implementation increases the sustainability and portability of the platform.The implemented multimodal Augmented Reality-based platform for training of maintenance andassembly skills has been evaluated in industry and compared to traditional training methods. The resultsshow that the developed training platform and the pursued training strategies are very well suited forthe training of maintenance and assembly skills and enhance traditional training.With the presented framework we have overcome the problems sketched above. We are cheap interms of effort and costs for the training of maintenance and assembly skills and we improve its efficiencycompared with traditional training.Item Privacy and Security Assessment of Biometric Template Protection(Zhou, 2011-09-19) Zhou, XuebingBiometrics enables convenient authentication based on a person s physical or behavioral characteristics. In comparisonwith knowledge- or token-based methods, it links an identity directly to its owner. Furthermore, it can notbe forgotten or handed over easily. As biometric techniques have become more and more efficient and accurate,they are widely used in numerous areas. Among the most common application areas are physical and logicalaccess controls, border control, authentication in banking applications and biometric identification in forensics.In this growing field of biometric applications, concerns about privacy and security cannot be neglected. Theadvantages of biometrics can revert to the opposite easily. The potential misuse of biometric information is notlimited to the endangerment of user privacy, since biometric data potentially contain sensitive information likegender, race, state of health, etc. Different applications can be linked through unique biometric data. Additionally,identity theft is a severe threat to identity management, if revocation and reissuing of biometric referencesare practically impossible. Therefore, template protection techniques are developed to overcome these drawbacksand limitations of biometrics. Their advantage is the creation of multiple secure references from biometric data.These secure references are supposed to be unlinkable and non-invertible in order to achieve the desired level ofsecurity and to fulfill privacy requirements.The existing algorithms can be categorized into transformation-based approaches and biometric cryptosystems.The transformation-based approaches deploy different transformation or randomization functions, whilethe biometric cryptosystems construct secrets from biometric data. The integration in biometric systems is commonlyaccepted in research and their feasibility according to the recognition performance is proved. Despiteof the success of biometric template protection techniques, their security and privacy properties are investigatedonly limitedly.This predominant deficiency is addressed in this thesis and a systematic evaluation framework for biometrictemplate protection techniques is proposed and demonstrated:Firstly, three main protection goals are identified based on the review of the requirements on template protectiontechniques. The identified goals can be summarized as security, privacy protection ability and unlinkability.Furthermore, the definitions of privacy and security are given, which allow to quantify the computational complexityestimating a pre-image of a secure template and to measure the hardness of retrieving biometric datarespectively.Secondly, three threat models are identified as important prerequisites for the assessment. Threat modelsdefine the information about biometric data, system parameters and functions that can be accessed during theevaluation or an attack. The first threat model, so called naive model, assumes that an adversary has very limitedinformation about a system. In the second threat model, the advanced model, we apply Kerckhoffs principleand assume that essential details of algorithms as well as properties of biometric data are known. The last threatmodel assumes that an adversary owns large amount of biometric data and this allows him to exploit inaccuracyof biometric systems. It is called the collision threat model.Finally, a systematic framework for privacy and security assessment is proposed. Before an evaluation process,protection goals and threat models need to be clarified. Based on these, the metrics measuring different protectiongoals as well as an evaluation process determining the metrics will be developed. Both theoretical evaluation withmetrics such as entropy, mutual information and practical evaluation based on individual attacks can be used.Item GPU Data Structures for Graphics and Vision(Ziegler, 2011-05-06) Ziegler, GernotGraphics hardware has in recent years become increasingly programmable, and its programming APIs use the stream processor model to expose massive parallelization to the programmer. Unfortunately, the inherent restrictions of the stream processor model, used by the GPU in order to maintain high performance, often pose a problem in porting CPU algorithms for both video and volume processing to graphics hardware. Serial data dependencies which accelerate CPU processing are counterproductive for the data-parallel GPU. This thesis demonstrates new ways for tackling well-known problems of large scale video/volume analysis. In some instances, we enable processing on the restricted hardware model by reintroducing algorithms from early computer graphics research. On other occasions, we use newly discovered, hierarchical data structures to circumvent the random-access read/fixed write restriction that had previously kept sophisticated analysis algorithms from running solely on graphics hardware. For 3D processing, we apply known game graphics concepts such as mip-maps, projective texturing, and dependent texture lookups to show how video/volume processing can benefit algorithmically from being implemented in a graphics API. The novel GPU data structures provide drastically increased processing speed, and lift processing heavy operations to real-time performance levels, paving the way for new and interactive vision/graphics applications.