SVRecon: Sparse Voxel Rasterization for Surface Reconstruction

TL;DR

SVRecon tackles high-fidelity surface reconstruction by embedding Signed Distance Functions into the sparse voxel rasterization framework. The key idea is to overcome voxel-boundary discontinuities and local minima by using geometry-informed initialization from PI^3 and a spatial coherence mechanism across a hierarchical, multi-resolution voxel structure, coupled with NeuS-style SDF learning. The method achieves fast convergence and hole-free, high-quality surfaces on benchmarks like DTU and Tanks-and-Temples, outperforming naive SVRaster+SDF setups and yielding robust results across indoor and outdoor scenes. This approach offers a scalable, efficient pathway for accurate surface reconstruction using explicit sparse primitives with learned SDF fields.

Abstract

We extend the recently proposed sparse voxel rasterization paradigm to the task of high-fidelity surface reconstruction by integrating Signed Distance Function (SDF), named SVRecon. Unlike 3D Gaussians, sparse voxels are spatially disentangled from their neighbors and have sharp boundaries, which makes them prone to local minima during optimization. Although SDF values provide a naturally smooth and continuous geometric field, preserving this smoothness across independently parameterized sparse voxels is nontrivial. To address this challenge, we promote coherent and smooth voxel-wise structure through (1) robust geometric initialization using a visual geometry model and (2) a spatial smoothness loss that enforces coherent relationships across parent-child and sibling voxel groups. Extensive experiments across various benchmarks show that our method achieves strong reconstruction accuracy while having consistently speedy convergence. The code will be made public.

Figures (13)

• Figure 1: SVRecon. This paper introduces Signed Distance Function (SDF) on top of the recent sparse voxel rasterization framework (SVRaster) for surface reconstruction. From left to right, the figure shows results from SVRaster Sun_2025_CVPR_SVRaster, a naive combination of SVRaster and NeuS Wang_2021_NeurIPS_NeuS, and our method, SVRecon. While the naive extension exhibits significant high-frequency artifacts, our approach produces smooth and continuous surfaces. This improvement stems from addressing the lack of voxel-wise coherence in SVRaster.
• Figure 2: Our pipeline consists of two main stages: (a) Initialization and (b) Optimization. In (a), we initialize a coarse SDF from multi-view images by aligning PI$^3$ point maps wang2025pi3 at estimated camera poses into points at ground truth camera poses (\ref{['subsec:sparse_voxel_initialization_using_pointmap']}). In (b), we impose spatial coherence to ensure voxel continuity via inter-voxel association (\ref{['subsec:sparse_voxel_association']}), parent-child Eikonal/smoothness losses (\ref{['subsec:sdf_encoding_in_sparse_voxel']}), and progressive voxel allocation (\ref{['subsec:sdf_encoding_in_sparse_voxel']}). Finally, our model yields a clean final mesh in minutes.
• Figure 3: Coherent voxel structure. (a) illustrates child-child voxel regularization using fine cell unit. After $2^9$ resolution, all voxel corners are efficiently connected due to hierarchical regularization (b) and (c). Red voxel corners ($\bullet$) are associated through the continuity loss.
• Figure 4: Reconstructed mesh comparison on DTU. Four scenes (rows) and four methods (columns). From top to bottom, we show scans 40, 63, 65, and 114. Our method converges to cleaner and more complete geometry while preserving details (best viewed with zoom).
• Figure 5: Qualitative comparison on Tanks-and-Temples. Two scenes (rows) and three methods (columns). From top to bottom, we show scene Barn and Truck. Our method reconstructs clean geometry even in large-scale outdoor scenes. (best viewed with zoom).