3D Reconstruction
3D reconstruction from images is one of the central problems in computer vision (CV). In recent years, optimization-based approaches that enforce consistency between the estimated scene geometry and its observed appearance, such as differentiable rendering, NeuS, and 3D Gaussian splatting, have become widely adopted. However, significant technical challenges remain in estimating not only the shape of target objects but also their structural properties (e.g., branching patterns) and view-dependent appearance. We particularly focus on fine-grained 3D geometry, complex structures, and photorealistic appearance, aiming to develop foundational techniques for reconstructing challenging real-world objects.
Homogeneous Gaussian Splatting (HoGS)
Novel view synthesis has demonstrated impressive progress recently, with 3D Gaussian splatting (3DGS) offering efficient training time and photorealistic real-time rendering. However, reliance on Cartesian coordinates limits 3DGS’s performance on distant objects, which is important for reconstructing unbounded outdoor environments. We found that, despite its ultimate simplicity, using homogeneous coordinates, a concept on the projective geometry, for the 3DGS pipeline remarkably improves the rendering accuracies of distant objects. We therefore propose Homogeneous Gaussian Splatting (HoGS) incorporating homogeneous coordinates into the 3DGS framework, providing a unified representation for enhancing near and distant objects. HoGS effectively manages both expansive spatial positions and scales particularly in outdoor unbounded environments by adopting projective geometry principles. Experiments show that HoGS significantly enhances accuracy in reconstructing distant objects while maintaining high-quality rendering of nearby objects, along with fast training speed and real-time rendering capability.
The code is available here.
