SAMPLING AND ANTIALIASING OF DISCRETE 3-D VOLUME DENSITY TEXTURES

We propose two variants of a new image-space, front-to-back, constant-time filtering technique for sampling and anti-aliasing 3-D density voxel fields. Such voxel fields describe the distribution of a scalar variable (density) in 3-D space and are employed for visualizing several volume effects (haze, fog, smoke, clouds, etc.) used in realistic image synthesis, in flying, ship, and driving optical simulators, as well as in several cases of scientific visualization. First, we point out the problems which occur when such 3-D textures are sampled, especially when the point sampling Monte-Carlo method is used. 'Distance sampling' and 'pyramidal-volume sampling' are then introduced. The first technique samples the texture along a straight line defined by the eye position and the pixel midpoint, whereas the pyramidal-volume technique approximately samples the volume of the pyramid defined by the eye and the four pixel corners. In comparison to other existing methods, both methods greatly reduce aliasing and calculation time. Especially the second one provides a constant-time filtering, whereby minimizing the number of texture evaluations. We demonstrate the applicability of the proposed methods for animation as well as for visualization purposes.