In current computer graphics systems, virtual scenes are usually described internally as a set of geometric objects in a 3D coordinate system. These geometric descriptions are then transformed into a set of pixels that are displayed to the user on a monitor.
The process of transforming the 3D descriptions to a set of pixels is referred to as rendering. We have developed methods for assessing geometric errors introduced by the rendering process.
We have implemented methods for assessing geometric errors introduced by the rendering of three 3D geometric primitive forms: points, line segments, and triangles. These represent the simplest and by far the most commonly used geometric forms rendered in current computer graphics systems. We have developed error metrics based on the comparison the actual set of pixels rendered for a given 3D geometric primitive and an idealized representation of the rendering of that primitive. We have tested these error assessment methods on several commonly-used computing platforms and we present the results in the paper cited on this page.