Three-dimensional shape rendering from multiple images

A paradigm for automatic three-dimensional shape and geometry rendering from multiple images is introduced in this paper. In particular, non-photorealistic rendering (NPR) techniques in the style of pen-and-ink illustrations are addressed, while the underlying presented ideas can be used in other modalities, such as halftoning, as well. Existing NPR approaches can be categorized in two groups depending on the type of input they use: image based and object based. Using multiple images as input to the NPR scheme, we propose a novel hybrid model that simultaneously uses information from the image and object domains. The benefit not only comes from combining the features of each approach, it also minimizes the need for manual or user assisted tasks in extracting scene features and geometry, as employed in virtually all state-of-the-art NPR approaches. As particular examples we use input images from binocular stereo and multiple-light photometric stereo systems. From the image domain we extract the tonal information to be mimicked by the NPR synthesis algorithm, and from the object domain we extract the geometry, mainly principal directions, obtained from the image set without explicitly using 3D models, to convey shape to the drawings. We describe a particular implementation of such an hybrid system and present a number of automatically generated pen-and-ink style drawings. This work then shows how to use and extend well-developed techniques in computer vision to address fundamental problems in shape representation and rendering. (c) 2005 Elsevier Inc. All rights reserved.