LITA-GS: Illumination-Agnostic Novel View Synthesis via Reference-Free 3D Gaussian Splatting and Physical Priors

Han Zhou, Wei Dong †, Jun Chen, † Corresponding Author · 2025

Read LITA-GS: Illumination-Agnostic Novel View Synthesis via Reference-Free 3D Gaussian Splatting and Physical Priors by Han Zhou, Wei Dong †, Jun Chen, † Corresponding Author (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-00219.txt.

Description

Han Zhou, Wei Dong, and Jun Chen offered LITA-GS in 2025 for building 3D scenes from badly lit photographs. The popular Gaussian splatting method struggles in harsh or dim lighting: the initial point cloud recovered from such images misses detail, and noise, lost information. And color warping block a clean, usually exposed rebuild. Simply brightening each photo first also fails, because separate enhancement makes varied viewpoints disagree about lighting.

Their method needs no reference well-lit images. A first workflow extracts an actual prior that does not change with illumination, catching the scene's dependable spatial form. A form rendering plan built on that prior then renders views every time regardless of the first lighting. The design aims at faithful novel views of scenes photographed under adverse illumination, a common real-world condition.

Catalog record

CREATORHan Zhou, Wei Dong †, Jun Chen, † Corresponding Author
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-00219

Browse the full library · About the Tower of Babel