Search papers, labs, and topics across Lattice.
This paper introduces DirectFisheye-GS, a modification to the 3D Gaussian Splatting (3DGS) framework that allows for direct training on fisheye images without prior undistortion. They address the issue of floaters and artifacts at image edges, which arise from the increased distortion at the periphery of fisheye images and the original 3DGS's per-iteration random view selection. By introducing a feature-overlap-driven cross-view joint optimization strategy, they enforce geometric and photometric consistency across views, leading to improved reconstruction quality.
Fisheye cameras promise better 3D scene reconstruction with fewer inputs, but current methods bottleneck on pre-processing; DirectFisheye-GS skips the undistortion step and trains directly on fisheye images, unlocking the full potential of wide FOV capture.
3D Gaussian Splatting (3DGS) has enabled efficient 3D scene reconstruction from everyday images with real-time, high-fidelity rendering, greatly advancing VR/AR applications. Fisheye cameras, with their wider field of view (FOV), promise high-quality reconstructions from fewer inputs and have recently attracted much attention. However, since 3DGS relies on rasterization, most subsequent works involving fisheye camera inputs first undistort images before training, which introduces two problems: 1) Black borders at image edges cause information loss and negate the fisheye's large FOV advantage; 2) Undistortion's stretch-and-interpolate resampling spreads each pixel's value over a larger area, diluting detail density -- causes 3DGS overfitting these low-frequency zones, producing blur and floating artifacts. In this work, we integrate fisheye camera model into the original 3DGS framework, enabling native fisheye image input for training without preprocessing. Despite correct modeling, we observed that the reconstructed scenes still exhibit floaters at image edges: Distortion increases toward the periphery, and 3DGS's original per-iteration random-selecting-view optimization ignores the cross-view correlations of a Gaussian, leading to extreme shapes (e.g., oversized or elongated) that degrade reconstruction quality. To address this, we introduce a feature-overlap-driven cross-view joint optimization strategy that establishes consistent geometric and photometric constraints across views-a technique equally applicable to existing pinhole-camera-based pipelines. Our DirectFisheye-GS matches or surpasses state-of-the-art performance on public datasets.
Tsinghua AI