Neural Trajectory Fields for Dynamic Novel View Synthesis
This addresses the challenge of recreating time-varying sequences for applications in virtual reality or video editing, representing an incremental improvement over static scene methods.
The paper tackles the problem of synthesizing novel views for dynamic scenes from a limited set of photographs, introducing DCT-NeRF to learn smooth trajectories for points in space, resulting in high-quality reconstruction in dynamic regions.
Recent approaches to render photorealistic views from a limited set of photographs have pushed the boundaries of our interactions with pictures of static scenes. The ability to recreate moments, that is, time-varying sequences, is perhaps an even more interesting scenario, but it remains largely unsolved. We introduce DCT-NeRF, a coordinatebased neural representation for dynamic scenes. DCTNeRF learns smooth and stable trajectories over the input sequence for each point in space. This allows us to enforce consistency between any two frames in the sequence, which results in high quality reconstruction, particularly in dynamic regions.