No description available.

No description available.

Made available in DSpace on 2015-03-04T18:50:41Z (GMT). No. of bitstreams: 1 capitulo00.pdf: 149755 bytes, checksum: a2f94dd5a3a96753bd5a54659a575c98 (MD5) Previous issue date: 2006-04-07 / Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro / Main memory limitations can lower the performance of segmentation applications for large images or even make it undoable. In this work we integrate the T-Surfaces model and Out-of-Core isosurface generation methods in a general framework for segmentation of large image volumes. T-Surfaces is a parametric deformable model based on a triangulation of the image domain, a discrete surface model and an image threshold. Isosurface generation techniques have been implemented through an Out-of-Core method that uses a kd-tree structure, called Meta-Cell technique. By using the Meta-Cell framework, we present an Out-of-Core version of a segmentation method based on T-Surfaces and isosurface extraction. The Gradient Vector Flow (GVF) is an approach based on Partial Differential Equations. This method has been applied together with snake models for image segmentation through boundary extraction. The key idea is to use a diffusion-reaction PDE in order to generate a new external force field that makes snake models less sensitivity to initialization as well as improves the snake s ability to move into boundary concavities. In this work, we firstly review basic results about global optimization conditions of the GVF and numerical considerations of usual GVF schemes. Besides, we present an analytical analysis of the GVF and a frequency domain analysis, which gives elements to discuss the dependency from the parameter values. Also, we discuss the numerical solution of the GVF based in a SOR method. We observe that the model can be used for Multiply Connected Domains and applied an image processing approach in order to increase the GVF efficiency. / Limitações de memória principal podem diminuir a performance de aplicativos de segmentação de imagens para grandes volumes ou mesmo impedir seu funcionamento. Nesse trabalho nós integramos o modelo das T-Superfícies com um método de extração de iso-superfícies Out-of-Core formando um esquema de segmentação para imagens de grande volume. A T-Superficie é um modelo deformável paramétrico baseado em uma triangulação do domínio da imagem, um modelo discreto de superfície e um threshold da imagem. Técnicas de extração de isso-superfícies foram implementadas usando o método Out-of-Core que usa estruturas kd-tree, chamadas técnicas de Meta-Células. Usando essas técnicas, apresentamos uma versão Out-of-Core de um método de segmentação baseado nas T-Superfícies e em iso-superfícies. O fluxo do Vetor Gradiente (GVF) é um campo vetorial baseado em equações diferenciais parciais. Esse método é aplicado em conjunto com o modelo das Snakes para segmentação de imagens através de extração de contorno. A idéia principal é usar uma equação de difusão-reação para gerar um novo campo de força externa que deixa o modelo menos sensível a inicialização e melhora a habilidade das Snakes para extrair bordas com concavidades acentuadas. Nesse trabalho, primeiramente serão revistos resultados sobre condições de otimização global do GVF e feitas algumas considerações numéricas. Além disso, serão apresentadas uma análise analítica do GVF e uma análise no domínio da frequência, as quais oferecem elementos para discutir a dependência dos parâmetros do modelo. Ainda, será discutida a solução numérica do GVF baseada no método de SOR. Observamos também que o modelo pode ser estendido para Domínios Multiplamente Conexos e aplicamos uma metodologia de pré-processamento que pode tornar mais eficiente o método.

This thesis focuses on the design and implementation of a software system enabling out-of-core rendering of multiple map datasets mapped on virtual globes around our solar system. Challenges such as precision, accuracy, curvature and massive datasets were considered. The result is a globe visualization software using a chunked level of detail approach for rendering. The software can render texture layers of various sorts to aid in scientific visualization on top of height mapped geometry, yielding accurate visualizations rendered at interactive frame rates. The project was conducted at the American Museum of Natural History (AMNH), New York and serves the goal of implementing a planetary visualization software to aid in public presentations and bringing space science to the public. The work is part of the development of the software OpenSpace, which is the result of a collaboration between Linköping University, AMNH and the National Aeronautics and Space Administration (NASA) among others.

This report covers theory and comparison of techniques for rendering massive scale 3D geospa- tial planet data in a web browser. It also presents implementation details of a few of these tech- niques in WebGL and Javascript, using the Three.js [1] 3D library. The thesis project is part of the implementation of Unitea, a web based education platform for interactive astronomy visualizations. Unitea is a derivative of Uniview, which is a fulldome interactive simulation of the universe. A major part of this thesis is dedicated to the implementa- tion of Hierarchical Level of Detail (HLOD) modules for Three.js based on the theory presented by T. Ulrich [2] and later generalized by Cozzi and Ring [3]. HLOD techniques are dynamic level of detail algorithms that represent the surface of objects as accurately as possible from a certain viewing angle. By using space partitioning tree-structures, view based error metrics and culling techniques detailed representations of the objects (in this case planets) can be efficiently rendered in real-time. The modules developed provide a general-purpose library for rendering planets (or other spher- ical objects) with dynamic level of detail in Three.js. The library also features connections to online web map services (WMS) and tile services.