Rendering virtual objects with high dynamic range lighting extracted automatically from unordered photo collections

Abstract

In this paper, we present a novel technique for applying image-based lighting on virtual objects by constructing synthesized environment maps (SEM) automatically. In contrast to related work, we use unordered photo collections as input instead of a completely measured environmental lighting (so called light probe). Each image is interpreted as a spatial piece of a light probe. Thus, we accumulate all given pieces to a synthesized environment map by a given weight function. In addition, multiple coverage is used to increase the dynamic range. Finally, we show how this algorithm is applied for an interactive image-based Augmented Reality environment for virtual product presentations. © 2009 Springer-Verlag.