Abstract: We present an accelerated volume rendering algorithm that efficiently exploits empty regions in the data set to minimize processing time. The volume data, which is stored in the video card, is rendered as a set of view-perpendicular slices through the data set. The new algorithm renders only the nonempty areas of each slice, and does so in a manner that permits classification operations to be performed in less than a third of a second.

The algorithm uses "off-the-shelf" commodity graphics hardware and achieves high levels of interactivity in both data manipulation and data classification, a common but usually ignored aspect of volume rendering. The new algorithm was tested with a wide variety of data sets and yielded frame rates between 2.9 and 5.1 times faster than the standard unaccelerated algorithm. The algorithm is very simple to implement and should be attractive to developers of volume rendering software.