Speaker: Emanuele Grisenti
Abstract
Computing an approximation of the radiance field of a given scene requires balancing construction time, reconstruction quality, memory consumption, and query cost. In particular, for lighting design optimization, we are interested in generating a high-quality representation that can be rebuilt quickly as the lighting configuration changes. In this context, we will explore an approach to view-independent radiance caching based on radiance cascades, a novel method for sampling the radiance field of a scene that uses different spatial and directional sampling resolutions for different distance intervals of traced rays. We will compare the resulting implementation to an existing solution that uses light tracing to approximate outgoing radiance at mesh vertices and stores directional information using hemispherical harmonics, and with a more conventional approach based on light probes.