If 100 people in the U.S. had a combined total of $100, one person would have $40, four people would have $5, and 95 people would have 42 cents.
some of my work :
Frequency domain volume rendering is a technique which allows
projections of n3 sized volume data to be generated in
O(n2 log n) time. This is achieved by exploiting the
projection-slice theorem which states that a projection in spatial
domain can be achieved by extracting a slice in frequency domain. This
paper first gives an overview of the theoretical principles, whereas
it also concentrates on the Hartley transform as a means of changing
between spatial and frequency domain. The Hartley transform is more
suitable in this context as the more familiar Fourier transform, since
it generates real output for real data. As an extension, depth cueing
to improve the spatial perception is described. Finally, some results
achieved by an actual implementation are given.
Keywords: volume rendering, Fourier transform, Hartley transform,
projection-slice theorem, scientific visualization, depth cueing
Reconstruction of gradients from sampled data is a crucial task in
volume visualization. Gradients are used, for instance, as normals for
surface shading or for classification in standard ray casting
techniques. Using the ideal derivative reconstruction filter, which
can be derived from the ideal function reconstruction filter, is
impracticable because of its infinite extend. Simply truncating the
filter leads to problems due to discontinuities at the edges. To
overcome these problems several windows have been defined, which are
discussed in this paper with respect to gradient estimation.
Keywords: volume visualization, reconstruction, gradient,
filter, frequency response
Highly elaborated visualization techniques that are based on surfaces
often are independent from the origin of the surface data. For
re-using advanced visualization methods for surfaces of various kind,
we developed an abstract surface interrogation layer called SMURF. In
this paper we discuss the steps necessary to unify multiple types of
surfaces under a shared general purpose interface.
Keywords: visualization, surfaces
Scientific visualization often deals with three-dimensional data.
Either the data is inherently three-dimensional, e.g., flow simulation
data or medical data, or 3D is chosen as the biggest reasonable
projection space for high-dimensional data, e.g., dynamical systems or
databases. Since conventional displays usually are two-dimensional,
this data has to be transformed to two dimensions which leads to
information loss. Objects in front occlude others lying behind them,
lines appear to coalesce into a unified solid surface of indeterminate
depth. This paper gives a short overview of some approaches which try
to enhance the spatial impression of three-dimensional data displayed
as two-dimensional pictures.
Keywords: perception, visualization, line shading, transparent
surfaces
This paper describes two approaches to triangulate a simple polygon. Emphasis is on practical and easy to implement algorithms, especially the first algorithm is straightforward and intuitive but, however, quite efficient. Further, it does not require the sorting or the use of balanced tree structures. Its worst running time complexity is O(n²), but for special classes of polygons it runs in linear time. The second approach requires some more sophisticated concepts of computational geometry but yields a better worst running time complexity of O(n log n). Both algorithms do not introduce new vertices and triangulate in a greedy fashion, that is they never remove edges once inserted. Further, they are designed to find an arbitrary triangulation and they do not optimize the result in any way.
Keywords: computational geometry, polygon, triangulation, computational complexity, monotone polygon, trapezoidation
Texture mapping is an efficient way to gain realism in computer generated pictures. However, if the texture is point sampled, aliasing occurs. Mip-Mapping is a technique which pre-filters the texture several times to successively lower resolutions. The adequate texture-level is then chosen depending on the distance of the point in three-space from the image plane. This reduces aliasing but leads to higher computational complexity. Both approaches were implemented and compared in a real-time application, namely NotAScratch, a three-dimensional computer action game.
theussl@cg.tuwien.ac.at[an error occurred while processing this directive]