Irregular and optimized representations for image based rendering

Harvard University, Cambridge MA

Investigators

Abstract

ABSTRACT 0203932 Gortler, Steven Harvard University Image based rendering is a promising area of innovation in computer graphics. In image based rendering, one starts with an input set of images or a video sequence. One then applies some degree of processing to this data to obtain an intermediate representation. This intermediate representation is then used in an interactive computer graphics setting. These interactive uses include but are not limited to: viewing the scene from novel points of view that were not available in the input sequence, relighting the scene using virtual lights that were not available in the input sequence, and geometrically manipulating the objects that make up the scene. This functionality greatly enlarges the uses of digital photographs, and introduces a rich source of content for use in interactive computer graphics. For image based rendering, we plan to focus on lumigraph representations, which are a sampled repre-sentation of the light that travels along the rays in the scene. In particular, the visual information of a static scene is represented by the five dimensional radiance function. This function describes the intensity and color of light for every 2D direction at every 3D spatial position (x; y; z). If we consider only the subset of light leaving the convex hull bounded object (or equivalently entering a bounded empty region of space), the fact that radiance along any ray remains constant 1 allows us to reduce the domain of interest of the radiance function from five to four dimensions.