Revisiting Photometric Ambiguity for Accurate Gaussian-Splatting Surface Reconstruction

Jiahe Li, Jiawei Zhang, Xiao Bai, Jin Zheng, Xiaohan Yu, Lin Gu, Gim Hee Lee

TL;DR: This paper presents AmbiSuR, a framework that explores an intrinsic solution upon Gaussian Splatting for the photometric ambiguity-robust surface reconstruction with high performance.

Abstract: Surface reconstruction with differentiable rendering has achieved impressive performance in recent years, yet the pervasive photometric ambiguities have strictly bottlenecked existing approaches. This paper presents AmbiSuR, a framework that explores an intrinsic solution upon Gaussian Splatting for the photometric ambiguity-robust surface reconstruction with high performance. Started by revisiting the foundation, our investigation uncovers two built-in primitive-wise ambiguities in representation, while revealing an intrinsic potential for ambiguity self-indication in Gaussian Splatting. Stemming from these, a photometric disambiguation is first introduced, constraining ill-posed geometry solution for definite surface formation. Then, we propose an ambiguity indication module that unleashes the self-indication potential to identify and further guide correcting underconstrained reconstructions. Extensive experiments demonstrate our superior performance in surface reconstruction compared to existing methods across various challenging scenarios, while excelling in broad compatibility. Our code will be made open-source upon acceptance.

Lay Summary:

When computers try to build accurate 3D models from standard 2D photos, they are often tricked by complex lighting, shiny reflections, or featureless textures. This confusion, where an "one-to-many" mapping problem makes different candidate 3D shapes can look exactly the same in photos, makes it incredibly hard for AI to figure out the true surface of an object.

In this paper, we introduce AmbiSuR, a new system designed to clear up this visual confusion. We dug into a popular 3D representation technique (Gaussian Splatting) and discovered something surprising: the system actually leaves hidden mathematical clues when it is "unsure" about a shape. AmbiSuR taps into this hidden potential, essentially giving the AI the ability to realize when it is making a bad guess. By catching these moments of confusion, our system automatically corrects itself to build solid, highly accurate 3D surfaces. Our tests show that AmbiSuR creates much more reliable and precise 3D models than previous methods, especially in visually tricky environments.

Link To Code: https://github.com/Fictionarry/AmbiSuR

Primary Area: Applications->Computer Vision

Keywords: 3D reconstruction, Gaussian Splatting, 3D vision

Submission Number: 4957