Multi-scale Fourier network for all-in-one adverse weather removal of power line aerial image

High-quality UAV-based aerial imagery is essential for reliable autonomous power line inspection. However, adverse weather conditions, such as haze, rain and snow, often degrade image clarity and hinder downstream analysis. This paper presents a Multi-scale Fourier Network (MFNET), an efficient all-in-one restoration framework designed to handle diverse weather degradations. The architecture integrates two primary modules: the Dual-branch Fourier Transform Block (DFTB) and the Multi-scale Structure (MSS). Specifically, the DFTB facilitates global feature extraction by independently calibrating amplitude and phase components in the frequency domain, optimizing the balance between representation capacity and computational overhead. Simultaneously, the MSS employs a cross-attention mechanism to fuse multi-scale features, enabling the model to capture hierarchical information across varying degradation levels. This integration effectively decouples entangled features inherent in unified restoration tasks. Quantitative and qualitative evaluations on multiple benchmarks demonstrate that MFNET achieves state-of-the-art performance and significantly enhances the robustness of power line instance segmentation.