Project is mainly focused on the reconstruction of damaged images while the stress is put on the finding of method giving the best possible visual results. We introduced a new algorithm and compared it with other existing methods (especially with method proposed by Ing. Karel Uhlíř). Investigated methods were tested on non-trivial real data damaged by various defect characters.

The project is concerned with clustering geometric data in a streaming fashion. Since geometric models are getting too large to fit into the main memory, they must be processed out-of-core. This means that the algorithm must work with a slow external memory. Random access is thus extremely inefficient and a data stream approach comes into account. Data stream clustering works in a hierarchical way which makes it possible to control the use of memory. Clustering can significantly reduce the amount of data. Additionally, the hierarchy can be exploited to create a multiresolution model.

This project focuses on compression and simplification of surface animations, i.e. dynamic meshes. We have so far published a state of the art report and an algorithm that seems to outperform current alternative approaches. We're using MVE-2 to develop the algorithms and testing tools, and we plan to publish some of the modules here.

This work foremost offers aids and techniques for development of creativity. Trying to key up humans to constructive thinking. It teaches some process during the solving of the problems.

.NET is a comfortable and easy-to-use environment for developing application that are not time-critical. However, it lacks official accelerated 3D output that would be cross-platform. The only official interface for .NET (CLI) platform is Direct3D. This project is an attempt to create Direct3D based interface that provides 3D accelerated and simple GUI and thus it runs on CLI implementations such as Mono, SSCLI.

Delaunay triangulation is one of the fundamental topics in the computational geometry and it is used in many areas, such as terrain modeling (GIS), scientific data visualization and interpolation, robotics, pattern recognition, meshing for finite element methods (FEM), natural sciences, computer graphics and multimedia, etc. Current applications often need to work with data sets such that they cannot be computed in one piece because of the common memory size limitations or their processing consumes too much time. In such cases, a parallel algorithm is useful and welcome. Quite a big set of parallel algorithms exists, however, most of these parallel algorithms were designed in times when parallel architectures, with hundreds of processors, dominated in the research area and thus they put stress usually on the scalability rather than on the robustness and simplicity. The main goal of this research is to design a robust and simple parallel algorithm suitable for architectures with several processors and the shared memory as well as for cluster architectures

The surface reconstruction is a common problem in computer graphics. Given a set of points sampled from some surface, a triangle mesh interpolating or approximating the points is to be obtained. As a typical real data set, e.g., points scanned on a statue, contains several millions of points, the surface reconstruction of such a data set is not an easy task because it demands a lot of memory (very often more than is available on a single computer). The goal of this research is to design parallel algorithm for clusters of workstations that can process such data sets due to the distributed storage and computing.

The main goal of this work is to evaluate suitability of subdivision surfaces for their application in interactive mesh modeling. Great attention is paid to natural ability of subdivision surfaces to represent multiresolution hierarchy of geometric details, and use of this property during mesh manipulation on different levels of details. The modeling tool built as a part of this work offers a basic set of editing operations, including simple copying of geometric details that were previously added to the model and their consecutive pasting to selected parts of the surface.

At present, many devices produce the volume data as their output. The well-known data acquisition tools are e.g. MRI, CT or PET scanners, which are used mainly in the medical field. The volume data can represent density, velocity, humidity etc. in the form of structured or unstructured grid. Scientists need to explore such data to study their inner properties and relations and to make decisions. The volume data visualization is a strong tool for such explorations. There are many methods for the volume data visualization. The volume rendering and the surface rendering approaches are governing two main visualization branches in this field. The volume rendering methods visualize the data as a whole. In the apart of that, the surface rendering methods are trying to find surfaces in the volume data and visualize them. The group of methods for the surface rendering contains also methods for the iso-surface extraction from the volume data. This project provides a reasonable overview of the main techniques that serve for the volume data acquisition as well as the state of the art in the field of the iso-surface extraction methods. The important methods are sensibly described together with appropriate references given and their difficulties, pros and cons are discussed as well. The offered work also concerns a research that was recently done and the outline of possible directions in which the future research can head towards.

General System for Distribution (GSD) is a system that enables users to create their own "distributed applications" without the need to have any further knowledge of distributed computing.

This work focus on Level of Detail terrain models. There is a large number of algorithms for rendering simplified representations of terrain. Algorithm known as ROAM was chosen and implemented with a few modifications.

A morphing or metamorphosis is the process of continuously transforming one geometric object into another. This technique is used mainly in animation (special effects) and design. Our work aims at 2D & 3D geometric objects given in boundary representation. The goal of our work is to develop methods which work more or less automatically (i.e. there is no need to manually specify correspondence between the source and target objects).

The MVE is a software tool that provides visual programming environment in which the user can manipulate data and create various graphical representations. It provides a framework for a module design.

Modular Visualization Environment 2 (MVE-2) project is based on .NET Technology. It offers easy-to-use, data-flow based modular environment. Its primary "customers" are researches and students together with their projects. Our environment makes these projects compatible with each other by minimum additional effort. API of module is forcing to follow good programming habits, such as clear problem decomposition, precise comment writing and cooperation with other programmers. Use of MVE-2 leads to less routine programming due to compatibility and reusability of existing modules. Therefore, users can concentrate on their particular problem and employ modules for marginal tasks.

Synthesizing realistic natural scenes is a challenging and important problem in computer graphics. The challenge stems not only from the complexity and diversity of objects and natural phenomena interacting together, but also from the huge amount of small details that should be modeled to obtain realistic models and physically plausible simulations. Therefore, there is still a need for interactive modeling and simulation techniques capable for handling complex synthetic sceneries. The purpose of this section is to explore novel techniques for modeling, animating, simulating and rendering complex natural phenomena.

desctiption

desctiption

desctiption

OpenGL is well known standard for low level rendering of basic graphics primitives used in wide range of applications. .NET environment offers user comfort and safety. This project is an attempt to provide benefits of both of the to user. The results will be OpenGL interface implementation in managed environment (.NET and other CLI implementations).

Although there are many methods solving the task of finding a path between two given spots, they usually assume the examined space does not change and is completely known in advance. Today's applications, however, do not have to meet these requirements, especially in case of virtual reality or computer games. Therefore, this project proposes a general model for the real-time path planning in a dynamic environment. It uses a regular triangulation for dynamic space subdivision and a heuristic algorithm providing fast way to find a path among all obstacles registered in the triangulation.

Solutions for the common problem of path planning in an abstract environment have been extensively developed in many scientific disciplines. However, almost all explored techniques assume the environment does not change and that there is a complete and detailed overview of this examined space. Moreover, many methods for the path planning need to derive a specific graph structure from the environment representation and it can be often very difficult to construct or obtain this structure in some real applications. We propose a general model for the real-time path planning in a known, partially known, unknown and dynamic environment. We provide a hybrid technique that combines a graph and grid representation of the examined space and that is trying to combine the advantages from both types of the environment representation.

Rendered stereoscopic scenes suffer from a scaling problem, which causes major restrictions on their portability to other stereoscopic display mediums. A main solution for this problem is presented, as well as other complementary solutions. Also, the basics of depth perception are presented, and some insight on the current three-dimensional displays used.

Shared virtual systems allow to remote users the more efficient cooperation on a common problem than it would be possible with tranditional means only (e.g., e-mail). The central part of these system is the transmission of users reprezentation (including also transmission of their actions or even emotions). Existing systems very often uses only a simple representation, thus users have limited possibilities of cooperation. The main goal of this project is to investigate the possibility of transmission of real scenes in the stereoscopic fashion, which would undoubtedly improve users options, and to design and implement system for sharing an artifictial scene among users allowing them to interact with objects in the scene.

In this work we studied a new solve to terrain simulation. Our solution consists in simulation of the terrain as a mesh. The mesh consists of points, which represent the height of terrain. Points can be inserted and deleted from the mesh. We can deform the mesh by virtual tools and the mesh is influenced by gravitation erosion, which causes gradual recovery of the displaced material.

Sort-Last is a computer graphics technique for rendering extremely large data sets on clusters of computers. Sort-Last works by dividing the data set into even-sized chunks for parallel rendering and then composing the images to form the final result. A subimage is generated by each processing unit and the final image is obtained by compositing subimages in the proper order. Sort-last rendering requires the movement of large amounts of image data among cluster nodes. The network interconnecting the nodes becomes a major bottleneck. Sort-last system was implemented as a easy to use library. Experiments shows that performance is mainly affected by the image composition time, not the transmission rate of network.

This thesis studies methods, which can be used to the construction of a partially dynamic regular triangulation. In the second part, tunnels in protein molecules and the application of regular triangulation for their searching are described. The tunnel's analysis is used in protein engineering for the research and modifications of protein molecule behavior.

Adjusted application is Kitware ParaView 3.3.0 development version. We adjusted it in Microsoft Visual Studio 2005 version 8.0.50727.762 (SP.050727-7600). Microsoft .NET Framework version 2.0.50727 running under Microsoft Windows XP Professional version 2002 including Service Pack 3. New control icons have been added into GUI of program. Which allows user to switch and operate stereoscopic mode by opening new instance of rendering window. Functionality of ParaView has been extended by possibility to operate on two monitors. It is possible to generate interlaced image of scene and adjust some of its settings, such as eye distance and go to full screen mode while leaving legends of image/scene non-interlaced and readable. All newly created windows have their own interactor.

Sublinear-time algorithms are more effective than the standard (linear, quadratic,..) algorithms, but they are often more difficult to implemen and sometimes they can produce an approximated solution only. This research is focused on both the accurate sublinear-time algorithms and approximations in sublinear time in context of the point location using walking algorithm in triangulations and the computation of convex hull.

This project attempts to improve quality of images gained by digital still cameras. Multiple images of the scene are taken and a single one of a higher resolution is reconstructed by improved Super-Resolution method.

The aim of this project is to provide and utilise approaches for visualisation of creatures. We'd like to fetch the creatures out of their squared prison and make them more realistic by giving them the life, even on the screen.

This project is devoted to the development of application that will provide a user-friendly visualization of 2D triangulations and compute various statistics (e.g., histogram of edge lengths, of minimal and maximal angles in triangles, etc.) in the selected parts of the triangulation.

Video representation is one of the most discussed topics within present multimedia. Despite of popular methods transcoding video in frequency domain, there is still need for geometric representations. First reason is an ability to perform temporal and spatial transformations without need for full recompression. Second reason is in exploring a new way of video coding. We proposed and implemented novel video compression scheme based on Delaunay triangulation. Such a triangulation is constructed from selected set of relevant points in actual frame. Its movement as motion compensation follows in selected number of consequent frames. Motion vectors are obtained by methods of motion estimation.

The aim of this work is straightforward use of visualisation toolkit on .NET platform. This is done by means of .NET wrappers. Each VTK class has its own wrapper. The wrapper contains wrapped Win32 class and provides access to its methods. Data conversion and memory management is also job for wrapper.