Information theory tools for viewpoint selection, mesh saliency and geometry simplification

Abstract

In this chapter we review the use of an information channel as a unified framework for viewpoint selection, mesh saliency and geometry simplification. Taking the viewpoint distribution as input and object mesh polygons as output vectors, the channel is given by the projected areas of the polygons over the different viewpoints. From this channel, viewpoint entropy and viewpoint mutual information can be defined in a natural way. Reversing this channel, polygonal mutual information is obtained, which is interpreted as an ambient occlusion-like quantity, and from the variation of this polygonal mutual information mesh saliency is defined. Viewpoint entropy, viewpoint Kullback-Leibler distance, and viewpoint mutual information are then applied to mesh simplification, and shown to compare well with a classical geometrical simplification method.