Efficient Dual-domain Image Dehazing with Haze Prior Perception

TL;DR

The Dark Channel Guided Frequency-aware Dehazing Network (DGFDNet), a dual-domain framework that explicitly aligns degradation across spatial and frequency domains, is proposed and achieves state-of-the-art performance with improved robustness and real-time efficiency.

Abstract

Transformers offer strong global modeling for single-image dehazing but come with high computational costs. Most methods rely on spatial features to capture long-range dependencies, making them less effective under complex haze conditions. Although some integrate frequency-domain cues, weak coupling between spatial and frequency branches limits their performance. To address these issues, we propose the Dark Channel Guided Frequency-aware Dehazing Network (DGFDNet), a dual-domain framework that explicitly aligns degradation across spatial and frequency domains. At its core, the DGFDBlock consists of two key modules: 1) Haze-Aware Frequency Modulator (HAFM), which uses dark channel priors to generate a haze confidence map for adaptive frequency modulation, achieving global degradation-aware spectral filtering. 2) Multi-level Gating Aggregation Module (MGAM), which fuses multi-scale features via multi-scale convolutions and a hybrid gating mechanism to recover fine-grained structures. Additionally, the Prior Correction Guidance Branch (PCGB) incorporates feedback for iterative refinement of the prior, improving haze localization accuracy, particularly in outdoor scenes. Extensive experiments on four benchmark datasets demonstrate that DGFDNet achieves state-of-the-art performance with improved robustness and real-time efficiency. Code is available at: https://github.com/Dilizlr/DGFDNet.

Figures (9)

• Figure 1: Comparison results on the SOTS-Indoor dataset. The bubble size represents the number of model parameters, and the number below each model indicates the SSIM value.
• Figure 2: (a) Investigation of the degradation characteristics of hazy images in the frequency domain by separately swapping the phase and imaginary components of hazy and clean images. (b) Demonstration of the global impact of frequency domain operations in the spatial domain by modifying local regions of the amplitude in hazy images.
• Figure 3: Overview of DGFDNet. (a) It includes a Dehazing Main Branch and Prior Correction Guidance Branch (PCGB), both with a three-scale symmetric design and inter-stage information exchange. The core module, DGFDBlock, incorporates (b) HAFM for global context modeling and (c) MGAM for local information modeling.
• Figure 4: The detailed structure of PCGB. PCGB follows the three-scale design of the dehazing main branch, progressively fusing original and feedback dark channel features through SKFusion, which is shared across all DGFDBlocks at each stage.
• Figure 5: Visual comparisons on synthetic hazy images from the SOTS-Indoor dataset. Key regions highlighted by red boxes are enlarged in the lower-left corner for clearer comparison.