Abstract
Reconstructing high-fidelity surfaces of dynamic scenes remains a critical challenge. While current
methods excel at novel view synthesis, they often struggle to recover accurate and stable geometry,
particularly under large non-rigid deformations. This results in noisy meshes that are unsuitable for
downstream simulation or editing.
In this work, we introduce a new paradigm for dynamic surface reconstruction based on a parametric
Signed Distance Function (p-SDF). Our key insight is to generalize static SDF fields—where
each spatial point stores a constant value—into time-dependent parametric curves where each curve
models a temporally evolving SDF trajectory. This parametric SDF representation provides a principled way
to capture complex temporal variations, naturally enforcing smoothness and continuity in shape dynamics.
At each timestamp, a static SDF field can be queried from p-SDF and converted into an explicit surface
mesh via differentiable iso-surfacing. By rendering these meshes with a physically based differentiable
renderer, we optimize the underlying parametric curves end-to-end against 2D image observations. Our
framework produces high-fidelity surfaces and inherently disentangles geometry, material, and lighting
from multi-view videos. It robustly reconstructs geometry under large-scale motions and resolves
appearance ambiguities caused by lighting and occlusions.
Experiments on both synthetic and real-world scenes demonstrate that our method achieves state-of-the-art
geometric accuracy.
