Reducing polygonal data by structural grouping algorithm

Abstract

In the field of computer graphics, polygonal representations are used for modeling three-dimensional geometrical objects. When recognizing structural characteristics, however, they often have much redundancy. Large numbers of polygons are difficult to render on graphic workstations or transfer over the network. To allow remote handling of polygonal data in virtual reality environments, the rational reduction of polygonal data is required. This paper describes a new algorithm which reduces the number of such polygonal primitives without losing the detailed structures of an object. This method is a kind of grouping algorithm and is effective for reducing structural redundancy. In the reducing process, we merge adjoining polygons, which satisfy a given coplanar criterion, into one plane. This grouping method is designed not to destroy the object structural patterns. The geometrical data are divided into groups satisfying the condition of required accuracy. The local reduction rule of polygons is applied to each classified group. We test this reducing method with geometric models representing human faces. The effectiveness for reducing polygon numbers and keeping the 3-D rendered image quality is investigated.