Focus on Scientific Visualization

This chapter presents an overview of techniques for visualization of fluid flow data. As a starting point, a brief introduction to experimental flow visualization is given. The rest of the chapter concentrates on computer graphics flow visualization. A pipeline model of the flow visualization process is used as a basis for presentation. Conceptually, this process centres around visualization mapping, or the translation of physical flow parameters to visual representations. Starting from a set of standard mappings partly based on equivalents from experimental visualization, a number of data preparation techniques is described, to prepare the flow data for visualization. Next, a number of perceptual effects and rendering techniques are described, and some problems in visual presentation are discussed. The chapter ends with some concluding remarks and suggestions for future development.

Three-dimensional visualization of medical objects from tomographic volume data is increasingly considered useful in various fields. This paper reviews methods for all steps of the 3D imaging pipeline from data preprocessing to object definition and display, with Special emphasis on advanced segmentation methods and surface- and voxel-based rendering techniques. Furthermore, multimodality matching, data manipulation, and aspects of image fidelity and implementation are discussed. Methods are illustrated with applications in craniofacial surgery, traumatology, neurosurgery, radiotherapy, and medical education.

Like in other disciplines, visualization is a fundamental method for environmental protection tasks. Researchers, practitioners, politicians and the public use graphics to get deeper insights into environmental problems and processes. As users, data, processes and models differ a lot, many kinds of visualization and user interface techniques are used.

This chapter presents a collection and compilation of informations about how scientific data is structured under the requirements of storing, processing and visualizing these data. Many different Software packages from different application areas as well as scientific visualization packages are available. While they differ in many points, the commonalities in describing and structuring data are dominant.

There is a recent increase in the use of discrete voxel representation for a variety of geometry-based applications. These include CAD, simulation, and scientific visualization, as well as those that intermix geometric objects with 3D sampled or computed datasets. In these applications the inherently continuous 3D geometric scene is sampled employing voxelization (3D scan-conversion) algorithms, which generate a 3D raster of voxels. The voxelized objects have to conform to some 3D discrete topological requirements such as connectivity and absence of tunnels. During the voxelization process, also termed the volume synthesis process, each voxel is assigned precomputed numeric values that represent some measurable view-independent properties of a tiny cube of the real or simulated object. These values are then readily accessible for speeding up the rendering or the discrete ray tracing process. The voxelization algorithms are the 3D counterparts of the 2D scan-conversion algorithms, and the 3D raster generated by them is the 3D counterpart of the conventional 2D raster.

An essential technique of visualizing the inner structure of a body is tomography. Tomographic methods represent the body by a sequence of cross sectional slices. For the observer, however, it is often rather difficult to imagine the true three-dimensional shape from the slices. For this reason, numerous methods for reconstructing a three-dimensional representation from a given stack of slices were suggested in the past. We give a unifying survey of such methods and sketch several new approaches not published yet. The emphasis is on techniques which can be also applied to larger distances between the cross sections and in the case of distorted slices. The methods usually yield a surface representation which can be used beyond visualization for purposes of geometric modeling, simulation, and manufacturing.

An introduction and overview of methods for modeling and visualizing trivariate data is presented. One variable is identified as being dependent on three other, independent variables. Data of this type arises often in practical problems of science and engineering The modeling portion is concerned with finding a mathematical relationship which represents the data. Visualization is concerned with using Computer generated images to convey information so that the user can learn about the relationship. The discussion on modeling methods will concentrate on the two application areas of surface-on-surface and Volumetric data. Volumetric data Covers the case where the independent data values represent points in a three dimensional domain. The surface-on-surface case is where the independent data is restricted to lie on some surface such as the earth or the surface of an airplane wing. Mathematical models with volume or surface domains are developed, discussed and compared. Graphing and visualizing trivariate relationships is quite challenging. Extending methods which have proven to be successful in other situations is a starting point. Several resulting modifications and extensions for both the case of surface-on-surface and Volumetric data are presented, discussed and compared. This includes some interactive techniques, isosurface algorithms, volume rendering and hypersurface projection graphs. Some new interrogation methods are also presented.

Free-form curves and surfaces are very important for sophisticated CAD/CAM Systems. Apart from the geometric modelling aspect of these curves and surfaces, the analysis of their quality is a necessary tool in the design and construction process. The purpose of this paper is to give a critical survey on curve and surface interrogation methods and to present generalized focal surfaces as a new surface interrogation tool.

There are basically two ways to visualize a scalar function in a volume: (1) draw contour surfaces, or (2) integrate a continuous volume density along viewing rays. The polyhedron compositing scheme of Max, Hanrahan, and Crawfis [9] combines both of these techniques by subdividing the volume cells at contour surfaces, and compositing the resulting polyhedral pieces and surface polygons in back-to-front order. The application of this algorithm to two specific situations is described, each using a special back-to-front sorting method. One application uses hierarchical data resulting from adaptive mesh refinement, and the other uses cloud data from climate simulations.

Different existing methods for the rendering of different classes of data, i.e. surface and volume data, in one image are described and discussed. Two fundamentally different strategies for combining the rendering of volume data with the rendering of surface or other geometric primitives have been developed. The first strategy is to convert either volume data into polygonal data or to convert polygonal data into volume data. Applying one of these conversions results in one class of data, which can be rendered with one appropriate rendering method. The second strategy is to use different rendering methods for the different classes of data and to combine the results in a final display. The rendering modules are either independent from each other or more or less tightly coupled. This includes merging the results of different rendering processes in the z-buffer of a workstation.

Modern medical imaging modalities, such as computed tomography (CT) and magnetic resonance imaging (MRI), produce sequences of cross-sectional images. They represent internal spatial structures of underlying medical features. Effective strategies for the direct display of the 3D information are essential to support the diagnostic process. In this chapter, we present a new display technique which allows fast volume rendering on low cost workstations.

The Fourier Projection Slice Theorem (FPST) is widely used for Medical Imaging, in particular for tomography. Here we develop a technique for using the FPST for display of volume data. Images are rendered from a frequency domain representation of the data, as opposed to the spatial representation which is more typical in Computer graphics. The principal advantage of this technique lies in the increase speed at which the algorithm runs compared to standard volume rendering algorithms. In fact, one can achieve a complexity of 1–3 orders of magnitude less than either a screen space or object space volume rendering techniques. In addition, assurances of image accuracy can be made. The principal drawback of the technique is the lack of hidden surface effects which makes the images more difficult to interpret compared with more conventional approaches. We present several techniques which are useful for ensuring artifact-free imagery. Among these are resampling filter design approaches, spatial premultiplication and zero padding.

This chapter presents an improved shading method especially useful for non-interactive radiosity based renderers. This method can be used to generate smooth realistic images while reducing the number of patches needed during the rendering process. It may be used for the visualization of radiosity scenes in combination with ray tracers, painters or depth-buffer systems to reduce Mach bands and other unwanted effects of linear interpolation.

Some extensions of the well-known ray tracing method are proposed which are used to produce images of objects inserted into one another with enhanced visual comprehensibility of shape and relative position of the 3-D objects by simulated radio diagnostics. These extensions are based on tracing a ray called X-ray to generate non-photo-realistic images - i.e. images beyond photo-realism - as well as photo-realistic images improved by making use of information that is usually redundant. The underlying geometric model is mainly based on surfaces, not on solids or voxels.

The representation of data in sound is emerging as a complement to data visualization. Several pilot studies over the past decade have proved the concept of auditory data representation; however, there has been little formal research to measure the effectiveness of auditory data representation techniques or to increase our understanding of how they work. Until quite recently, appropriate computing environments for research in this area simply did not exist. This situation is now improving and an increasing number of investigators are seeking solutions to the formidable problems posed by auditory data representation. This paper surveys the present state of the field and outlines the central problems which must be solved if the auditory data representation is to become a truly useful tool for data analysis and exploration.

This section contains the color illustration of the previous chapters. The references to literature refer to the bibliographies at the end of the respective chapter.