Efficient Real-World Dehazing via Physics-Inspired Global-Local Decoupling

2026-06-24

AI Summary

This work addresses the highly ill-posed nature of single-image dehazing in real-world scenarios, where spatially and spectrally varying scattering effects complicate restoration while demanding lightweight models with low latency. To this end, we propose PGL-Net, which embeds physics-inspired inductive bias without explicitly estimating physical parameters through a global–local decoupling strategy. Specifically, a Physics-inspired Affine Fusion (PAF) module enables cross-scale global contextual alignment, while a compact Degradation-Aware Modulation (DAM) block adaptively reconstructs local details. Extensive experiments demonstrate that our method achieves state-of-the-art performance across multiple real-world benchmarks. Notably, the Tiny variant surpasses the current best method, SGDN, by up to 2.6 dB in PSNR while reducing inference latency by over 10×, and significantly enhances downstream object detection accuracy.

Abstract

Real-world single image dehazing is highly ill-posed due to spatially and spectrally varying scattering, while practical deployment demands lightweight and low-latency models. Existing approaches either rely on fragile physical inversion under simplified assumptions or adopt heavy blind architectures unsuitable for edge deployment. To overcome these limitations, we propose PGL-Net (Physics-Inspired Global-Local Decoupling Network), a lightweight framework that incorporates physical inductive biases via operator-level emulation, avoiding explicit parameter estimation. It decouples dehazing into global distribution rectification and local structural refinement. A Physics-Inspired Affine Fusion (PAF) module performs globally conditioned alignment across hierarchical skip connections to compensate for haze-induced bias, while a compact Degradation-Aware Modulation (DAM) block adaptively restores spatially and spectrally variant details through dynamic feature modulation. Extensive experiments on multiple real-world benchmarks demonstrate that PGL-Net achieves state-of-the-art restoration quality with significantly reduced complexity. Compared with the recent SOTA SGDN, the Tiny variant (PGL-Net-T) improves PSNR by up to 2.6dB and consistently enhances downstream object detection accuracy, while achieving over a 10x reduction in inference latency. Code is publicly available at: https://github.com/sc-30-bit/PGL-Net.

Problem

• real-world dehazing
• single image dehazing
• lightweight model
• low-latency deployment
• spatially-varying scattering

Innovation

• Physics-Inspired Global-Local Decoupling
• Lightweight Dehazing Network
• Operator-Level Physical Emulation
• Dynamic Feature Modulation
• Real-World Image Restoration