An efficient image mapping method for 3D surface textures

Abstract

AbstractTo generate 3D computer graphics models, we need to obtain their surface geometrical data as well as texture data. This task is usually quite laborious. When real objects are on hand, there are scanners to acquire these data directly. This paper is concerned with an efficient method to acquire texture data for an object whose geometrical data are already known. We set up 320 viewpoints around an object and select a minimal number of viewpoints to cover all of its surfaces. Photographs are taken from the viewpoints. For the surfaces not in the photographs, new photographs are usually taken from viewpoints closer to the object. Lastly, those photographs are mapped exactly onto the object's surface. This method has the advantage of not requiring exact setting of viewpoint positions, since the actual camera positions are found by moving known geometrical data virtually. We conducted experiments with three models whose geometrical data were acquired through CT. © 2001 Scripta Technica, Syst Comp Jpn, 32(4): 14–24, 2001