OmniNeRF: Hybriding Omnidirectional Distance and Radiance fields for Neural Surface Reconstruction
This work improves 3D reconstruction precision for applications like Virtual Reality and Automatic Driving, representing an incremental advancement over existing NeRF-based methods.
The paper tackles the problem of blurred surfaces in 3D reconstruction from images by proposing OmniNeRF, a hybrid implicit field combining omnidirectional distance and radiance fields, which effectively addresses NeRF defects at surface edges and provides higher quality reconstruction results.
3D reconstruction from images has wide applications in Virtual Reality and Automatic Driving, where the precision requirement is very high. Ground-breaking research in the neural radiance field (NeRF) by utilizing Multi-Layer Perceptions has dramatically improved the representation quality of 3D objects. Some later studies improved NeRF by building truncated signed distance fields (TSDFs) but still suffer from the problem of blurred surfaces in 3D reconstruction. In this work, this surface ambiguity is addressed by proposing a novel way of 3D shape representation, OmniNeRF. It is based on training a hybrid implicit field of Omni-directional Distance Field (ODF) and neural radiance field, replacing the apparent density in NeRF with omnidirectional information. Moreover, we introduce additional supervision on the depth map to further improve reconstruction quality. The proposed method has been proven to effectively deal with NeRF defects at the edges of the surface reconstruction, providing higher quality 3D scene reconstruction results.