Learning Signed Distance Field for Multi-view Surface Reconstruction

TL;DR

This paper introduces a neural surface reconstruction method that combines stereo matching and feature consistency to improve multi-view 3D reconstruction of complex scenes without requiring object masks.

Contribution

It proposes a novel framework using signed distance fields and surface light fields, directly supervised by stereo geometry and optimized for feature consistency.

Findings

Outperforms previous methods on DTU, EPFL, and Tanks and Temples datasets.

Supports reconstruction of complex, concave, and wide-open scenes without masks.

Achieves more accurate mesh reconstructions compared to state-of-the-art approaches.

Abstract

Recent works on implicit neural representations have shown promising results for multi-view surface reconstruction. However, most approaches are limited to relatively simple geometries and usually require clean object masks for reconstructing complex and concave objects. In this work, we introduce a novel neural surface reconstruction framework that leverages the knowledge of stereo matching and feature consistency to optimize the implicit surface representation. More specifically, we apply a signed distance field (SDF) and a surface light field to represent the scene geometry and appearance respectively. The SDF is directly supervised by geometry from stereo matching, and is refined by optimizing the multi-view feature consistency and the fidelity of rendered images. Our method is able to improve the robustness of geometry estimation and support reconstruction of complex scene topologies. Extensive experiments have been conducted on DTU, EPFL and Tanks and Temples datasets. Compared to previous state-of-the-art methods, our method achieves better mesh reconstruction in wide open scenes without masks as input.