Omnidirectional Neural Radiance Field for Immersive Experience

Qiaoge Li, Itsuki Ueda, Chun Xie, Hidehiko Shishido, Itaru Kitahara

Abstract: This paper proposes a method using only RGB information from multiple captured panoramas to provide an immersive observing experience for real scenes. We generated an omnidirectional neural radiance field by adopting the Fibonacci sphere model for sampling rays and several optimized positional encoding approaches. We tested our method on synthetic and real scenes and achieved satisfying empirical performance. Our result makes the immersive continuous free-viewpoint experience possible.