Issues on Displaying 3D Data for Scientific Visualization Thomas Theußl theussl@cg.tuwien.ac.at Institute of Computer Graphics Vienna University of Technology Vienna, Austria

Conclusions

When visualizing three-dimensional data on two-dimensional displays there are several points that have to be considered.
Lines in 3D do not have a unique normal vector, integrating over all normal vectors cannot be done efficiently, so one vector has to be chosen to shade the lines accordingly, for shading very much increases the spatial impression of the resulting images. Illuminated stream lines, where the vector coplanar to the light and tangent vector is chosen as normal vector, give a much better impression of the vector field structure than flat shaded stream lines. However, care must be taken to select the seed points in such a way that the resulting images do not get overloaded and confusing. Detecting characteristic structures and depicting them properly with selected stream lines reduces occlusion and enhances the perception of the vector field structure.
Occlusion yet becomes a major problem when depicting surfaces. Transparency is useful to let one see through surfaces but impairs the spatial impression of the surface itself and the relative depth of multiple layered surfaces. The general approach is to use a sparse, opaque texture so the surface can be seen and seen through the same time. Standard hardware texture-mapping routines can be used to apply a opacity-modulating texture to a surface. This increases very much the perception of the shape of the surface itself but often impairs the perception of the inner structures. Other approaches try to emphasize characteristic parts of the surface, e.g., by depicting ridge and valley lines or by defining strokes dependent of the curvature of the surface. When depicting stream surfaces it is desirable to visualize the direction of the flow too what is nicely done by stream arrows, arrow-shaped textures mapped to the stream surfaces.