DehazeGS: Seeing Through Fog with 3D Gaussian Splatting

TL;DR

DehazeGS introduces an explicit Gaussian-based model for fog removal in multi-view images, enabling efficient and detailed scene reconstruction and rendering without relying on computationally intensive neural networks.

Contribution

The paper presents a novel Gaussian representation for fog modeling that improves dehazing accuracy and efficiency over existing neural network-based methods.

Findings

Achieves state-of-the-art dehazing performance on real and synthetic datasets.

Reconstructs detailed fog-free scenes from multi-view foggy images.

Operates efficiently by removing scattering effects directly from Gaussian primitives.

Abstract

Current novel view synthesis methods are typically designed for high-quality and clean input images. However, in foggy scenes, scattering and attenuation can significantly degrade the quality of rendering. Although NeRF-based dehazing approaches have been developed, their reliance on deep fully connected neural networks and per-ray sampling strategies leads to high computational costs. Furthermore, NeRF's implicit representation limits its ability to recover fine-grained details from hazy scenes. To overcome these limitations, we propose learning an explicit Gaussian representation to explain the formation mechanism of foggy images through a physically forward rendering process. Our method, DehazeGS, reconstructs and renders fog-free scenes using only multi-view foggy images as input. Specifically, based on the atmospheric scattering model, we simulate the formation of fog by establishing the transmission function directly onto Gaussian primitives via depth-to-transmission mapping. During training, we jointly learn the atmospheric light and scattering coefficients while optimizing the Gaussian representation of foggy scenes. At inference time, we remove the effects of scattering and attenuation in Gaussian distributions and directly render the scene to obtain dehazed views. Experiments on both real-world and synthetic foggy datasets demonstrate that DehazeGS achieves state-of-the-art performance. visualizations are available at https://dehazegs.github.io/