Fi-Gaussian: Frequency-Aware Implicit Gaussian Splatting for Single Image Dehazing

2026-06-14

AI Summary

Single-image dehazing is often hindered by the loss of high-frequency details and the difficulty of accurately modeling physical scattering processes. This work proposes a frequency-aware implicit Gaussian splatting network, introducing implicit Gaussian splatting—a technique previously unexplored in low-level vision tasks—to address these challenges. By continuously modeling the distribution of haze-free images in a 2D feature space, the method integrates frequency-domain disentanglement with complex-weighted Gaussian aggregation to recover fine details. Furthermore, it incorporates a physics-driven scattering renormalization mechanism that jointly estimates the transmission map and atmospheric light. This approach enables adaptive fusion of structural and textural information in the frequency domain and leverages implicit priors to enhance the accuracy of physical parameter estimation, achieving state-of-the-art quantitative performance and superior visual quality across multiple benchmark datasets.

Abstract

Single image dehazing continues to be hindered by the loss of high-frequency details and the difficulty of accurate physical scattering modeling. To address these issues, we propose Fi-Gaussian, a frequency-aware implicit Gaussian splatting network for single image dehazing. Unlike explicit rendering methods that rely on 3D point clouds, our method employs implicit Gaussian splatting to adaptively model the underlying distribution of clear images as a continuous representation in 2D feature space. The core of the network is a frequency-aware implicit Gaussian splatting module, which decouples low-frequency structural information and high-frequency texture information in the frequency domain and then performs adaptive Gaussian aggregation with complex-valued weights to recover fine details. In addition, a physics-driven scattering renormalization mechanism is introduced to estimate the transmission map and atmospheric light under the guidance of implicit Gaussian priors. Extensive experiments on multiple benchmark datasets demonstrate that Fi-Gaussian achieves state-of-the-art quantitative performance and produces visually superior dehazed results, validating the effectiveness of implicit Gaussian splatting for low-level vision tasks.

Problem

• single image dehazing
• high-frequency details
• physical scattering modeling
• frequency-aware
• implicit representation

Innovation

• Implicit Gaussian Splatting
• Frequency-Aware
• Single Image Dehazing
• Complex-Valued Aggregation
• Physics-Driven Renormalization