Multiperspective Rendering Using GLCs

With: Leonard McMillan

Abstract

We present a framework for the direct rendering of multiperspective images. We treat multiperspective imaging systems as devices for capturing smoothly varying set of rays, and we show that under an appropriate parametrization, multiperspective images can be characterized as continuous manifolds in ray space. We use a recently introduced class of General Linear Cameras (GLC), which describe all 2D linear subspaces of rays, as primitives for constructing multiperspective images. We show GLCs when constrained by an appropriate set of rules, can be laid out to tile the image plane and, hence, generate arbitrary
multiperspective renderings. Our framework can easily render a broad class of multiperspective images, such as multiperspective panoramas, neocubist style renderings, and faux-animations from still-life scenes. We also show a method to minimize distortions in multiperspective images by uniformly sampling rays on a sampling plane even when they do not share a common origin.

In proceedings of Eurographics Symposium on Rendering 2004. The paper can be downloaded here. 

See also a video illustrating how our framework works. Our work can also be used for image-based faux animation, as shown in the video below.