...on rendering

New Integration Kernels for Visualizing Vector and Tensor Fields Vector- and tensor-valued fields are a challenge to visualize because the multidimensional values at each point in a dataset have complicated interactions. We believe that novel visual metaphors using new types of subvoxel textures, fibers and streamlines will be useful.

Volume rendering moved beyond the hard surfaces that geometric representations implied and provided a more continuous representation of the underlying data. The essential difference between the traditional volume visualization and graphics algorithms and the algorithms for specific visual effects centers on an internal integration kernel and the scattering model for light. We will investigate the development of new integration kernels for visualizing vector and tensor fields. Part of the integration kernels will relate to the construction of subpixel semitransparent fields of streamlines, vortices and other structures, similar to Kajiya's subvoxel rendering of hair and fur. Other combinations will be less physically-based, including vector-valued level sets and gradients to create the appearance of surfaces and sheets within the vector and tensor data These new integration kernels are intended to give us a greater ability to see finer detail in our multidimensional datasets, by combining much of the vector and tensor data into nonlinear lighting models and other highly compact forms of visual representation.