Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields
This addresses a limitation in view synthesis for glossy surfaces, which is incremental as it builds on NeRF.
The paper tackled the problem of Neural Radiance Fields (NeRF) failing to accurately capture glossy surfaces by introducing Ref-NeRF, which replaces NeRF's view-dependent radiance parameterization with a structured representation of reflected radiance and scene properties, resulting in significantly improved realism and accuracy of specular reflections.
Neural Radiance Fields (NeRF) is a popular view synthesis technique that represents a scene as a continuous volumetric function, parameterized by multilayer perceptrons that provide the volume density and view-dependent emitted radiance at each location. While NeRF-based techniques excel at representing fine geometric structures with smoothly varying view-dependent appearance, they often fail to accurately capture and reproduce the appearance of glossy surfaces. We address this limitation by introducing Ref-NeRF, which replaces NeRF's parameterization of view-dependent outgoing radiance with a representation of reflected radiance and structures this function using a collection of spatially-varying scene properties. We show that together with a regularizer on normal vectors, our model significantly improves the realism and accuracy of specular reflections. Furthermore, we show that our model's internal representation of outgoing radiance is interpretable and useful for scene editing.