Image synthesis based on the recognition of 3‐D polyhedras

Abstract

Abstract A methodology of synthesizing images from the point of view of considering the object's recognizability is developed. The viewing parameters determining the synthesized image are decided based on the recognizability of the object. An object is identified through the determination of its position and orientation. In this paper, it is shown that we cannot distinguish topographically similar objects from a single perspective. Thus, the threshold i, determined by a scheme developed here for model i, is used to check whether a candidate model is accepted. The procedures of synthesizing an image also consist of determining the lighting conditions. The direction of light source is so chosen that |i‘•n k1 – i‘ • nk2 | ≥ (1/K), where i’ is the opposite direction of light source, nk 1 and nk 2 are the normals of two visible neighboring surfaces, and K is the number of edges between the visible surfaces of an object. After the direction of the light source is obtained, the intensity of the light source is determined such that the differences between two visible neighboring planes should be greater than a threshold. Finally, all the concepts presented are illustrated through several examples given in this paper. The result may be applied to the icon representation of an object which represents an execution process or a filename in a pictorial information system.