Dynamic radiosity on multi-resolution models

Abstract

Radiosity is widely applied in the areas of architectural and lighting design, 3D games, or outdoor scene illuminations. It is a view independent global illumination solution that produces a realistic image of the real world. However, most existing global illumination methods are computationally very expensive when handling dynamic scenes, in which more than one object is undergoing geometric transformation. Hence, they are not suitable for use in interactive environments. To solve this problem, we propose the Dynamic Radiosity method to provide an interactive update solution for radiosity, to handle dynamic environments with diverse conditions of dynamic objects. In our research, we engage multi‐resolution modeling into radiosity to accelerate computing the energy interactions among the objects. It mainly concerns with how to redistribute the patch radiosities computed at one resolution of the object model to another resolution, while guaranteeing the overall energy conservation. The framework and key techniques are illustrated in this thesis, and with the basic experiments, it is shown that dynamic radiosity is typically suitable for supporting realistic rendering in applications that require interactive rendering frame rate. Meanwhile, it is the first approach to handle scenes with simultaneous motions of multiple dynamic objects, even with invisible ones. It is achieved by the prioritized update decision that we have developed, which concerns with scene geometry and object attributes.