Addressing the Shape-Radiance Ambiguity in View-Dependent Radiance Fields

We present a method for handling view-dependent information in radiance fields to help with convergence and quality of 3D reconstruction. Radiance fields with view-dependence suffers from the so called shape-radiance ambiguity, which can lead to incorrect geometry given a high angular resolution of view-dependent colors. We propose the addition of a difference plane in front of each camera, with the purpose of separating view-dependent and Lambertian components during training. We also propose an additional step where we train, but do not store, a low-resolution view-dependent function that helps to isolate the surface if such a separation is proven difficult. These additions have a small impact on performance and memory usage but enables reconstruction of scenes with highly specular components without any other explicit handling of view-dependence such as Spherical Harmonics.

PERF: Performant, Explicit Radiance Fields

We present a novel way of approaching image-based 3D reconstruction based on radiance fields. The problem of volumetric reconstruction is formulated as a non-linear least-squares problem and solved explicitly without the use of neural networks. This enables the use of solvers with a higher rate of convergence than what is typically used for neural networks, and fewer iterations are required until convergence. The volume is represented using a grid of voxels, with the scene surrounded by a hierarchy of environment maps. This makes it possible to get clean reconstructions of 360{\deg} scenes where the foreground and background is separated. A number of synthetic and real scenes from well known benchmark-suites are successfully reconstructed with quality on par with state-of-the-art methods, but at significantly reduced reconstruction times.