Method for triangle subdivision in computer graphics texture mapping to eliminate artifacts in high perspective polygons

DETAILED DESCRIPTION The present invention relates to rendering three-dimensional graphics images on a computer controlled display system and more particularly to a system and method that eliminates artifacts created when large polygons with high perspective ratios are rendered/ texture mapped.
The present invention is useful within a computer controlled display system utilizing interpolation driven rendering/texture mapping processes.
The system and method of the present invention performs an iterative operation that subdivides selected polygons into a plurality of smaller polygons to limit artifact creation during the interpolation driven rendering/texture mapping processes.
Specifically, the processing logic of the present invention analyzes each polygon stored in display list memory of a graphics accelerator or graphics subsystem and determines a perspective ratio between adjacent vertices of the polygon.
If the perspective ratio is greater than a pre-selected limit, the edge bounded by the vertices is subdivided at the mid-point and new triangles created.
The process is iterative until all polygons have perspective ratios that are less than the pre-selected limit at which time the object data can be rasterized and displayed by the hardware.
Since the subdivision is based on determining the midpoint of edges rather than vertices of the polygon, any original polygon that shares the selected edge (and there might be several) will share a common vertex because the edge will be subdivided at the same mid-point thereby limiting creation of artifacts even if the edge appears later in the data stream.
Further, since the processing logic of the present invention consistently subdivides the edge each time it is encountered, objects are consistently rendered regardless of the coordinates of the vertices of the various polygons sharing the edge or having a co-incident edge.
The present invention calculates perspective values (w) and new coordinates (x, y, z) for each mid-point based on a simple average of the perspective values at the vertices bounding the edge