Geometry and Gradient-based Partitioning for Panoramic Outdoor Reconstruction
2026-07-09 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address the challenge of creating detailed 3D models of large outdoor areas using 360-degree panoramic images, which are hard to process with existing methods. They introduce PanoLOG, a two-step approach that first gets a rough 3D shape and then improves it using a special way to divide the scene based on geometry and gradients. They also created a new dataset called Pano360 for testing this kind of work. Their method helps improve 3D rendering quality while allowing efficient, smaller-scale training instead of handling everything at once.
3D Gaussian SplattingEquirectangular ProjectionPanoramic ImagesGeometry-based PartitioningGradient-based PartitioningSky-sphere ModelingMonocular Depth SupervisionParallaxOutdoor Scene ReconstructionBlock-parallel Training
Authors
Weijian Chen, Weibo Yao, Yuhang Zhang, Xiaolin Tang, Guo Wang, Weijun Zhang, Xitong Gao, Yihao Chen, Hongde Qin, Lu Qi
Abstract
Scaling 3D Gaussian Splatting (3DGS) to large outdoor scenes is costly in both data acquisition and computation. Adopting panoramic images with equirectangular projection (ERP) can reduce capture effort via their full $360^{\circ}$ field of view, yet the resulting omnipresent visibility invalidates existing partitioning strategies that rely on local camera frustums, causing block-wise optimization to degenerate into global training. Thus, we propose PanoLOG, a two-stage coarse-to-fine framework equipped with a Geometry and Gradient-based Partitioning Strategy tailored for large-scale panoramic 3DGS reconstruction. In the global coarse stage, PanoLOG leverages sky-sphere modeling and panoramic monocular depth supervision for reliable geometry, while in the refinement stage, G$^2$PS builds adaptive bounding volumes via parallax-driven uncertainty and assigns cameras via gradient-based importance scoring. Furthermore, we construct Pano360, the first benchmark on large-scale panoramic dataset for outdoor scene reconstruction. Extensive experiments demonstrate that G$^2$PS achieves state-of-the-art rendering quality while maintaining scalable, block-parallel training. Our models, training code, and dataset are publicly available.