FFA-Net: Feature Fusion Attention Network for Single Image Dehazing

By Xu Qin, Zhiling Wang, Yuanchao Bai, Xiaodong Xie, Huizhu Jia. AAAI Conference on Artificial Intelligence, 2019

1,948 citations recorded by the source corpus.

In this paper, we propose an end-to-end feature fusion at-tention network (FFA-Net) to directly restore the haze-free image. The FFA-Net architecture consists of three key components:1) A novel Feature Attention (FA) module combines Channel Attention with Pixel Attention mechanism, considering that different channel-wise features contain totally different weighted information and haze distribution is uneven on the different image pixels. FA treats different features and pixels unequally, which provides additional flexibility in dealing with different types of information, expanding the representational ability of CNNs. 2) A basic block structure consists of Local Residual Learning and Feature Attention, Local Residual Learning allowing the less important information such as thin haze region or low-frequency to be bypassed through multiple local residual connections, let main network architecture focus on more effective information. 3) An Attention-based different levels Feature Fusion (FFA) structure, the feature weights are adaptively learned from the Feature Attention (FA) module, giving more weight to important features. This structure can also retain the information of shallow layers and pass it into deep layers.The experimental results demonstrate that our proposed FFA-Net surpasses previous state-of-the-art single image dehazing methods by a very large margin both quantitatively and qualitatively, boosting the best published PSNR metric from 30.23 dB to 36.39 dB on the SOTS indoor test dataset. Code has been made available at GitHub.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/22418b7b7d5d6d18b81f232f60c22a15d1c8a38a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Camouflaged Object Detection with Feature Decomposition and Edge Reconstruction (2023, methodology): “We then employ the joint attention module JA (•), which includes spatial attention [34] and channel attention [45], to highlight noteworthy parts in both spatial and channel domains.”
Image-Adaptive YOLO for Object Detection in Adverse Weather Conditions (2021, neutral): “A straightforward approach is to pre-process the image by using the classical dehazing or image enhancement methods (Guo et al. 2020; He, Sun, and Tang 2009; Liu et al. 2019; Hang et al. 2020; Qin et al. 2020), which are originally designed to remove fog and enhance image quality.”
Vision Transformers for Single Image Dehazing (2022, methodology): “Comparing the baseline methods, FFA-Net and AECR-Net are far superior to the previous or contemporaneous methods.”
Omni-Kernel Network for Image Restoration (2024, neutral): “For instance, FFA-Net (Qin et al. 2020) utilizes the channel attention and pixel attention modules to treat channel-wise and pixel-wise features unequally for uneven haze distribution.”
MAXIM: Multi-Axis MLP for Image Processing (2022, background): “Visual comparisons for image dehazing on SOTS indoor testset [45] among GCANet [13], GridDehaze [54], DuRN [55], MSBDN [22], FFA-Net [72], and our MAXIM-2S.”
PSD: Principled Synthetic-to-Real Dehazing Guided by Physical Priors (2021, methodology): “FFA-Net [11].”
DEA-Net: Single Image Dehazing Based on Detail-Enhanced Convolution and Content-Guided Attention (2023, methodology): “Most of existing dehazing methods [5], [6], [12] adopt classical convolution layers for feature extraction without utilizing any priors.”
Image Dehazing Transformer with Transmission-Aware 3D Position Embedding (2022, methodology): “(f) FFANet [24] (g) MSBDN [9] (h) UHD [36] (i) DeHamer (ours) (j) GT”
Contrastive Learning for Compact Single Image Dehazing (2021, methodology): “Actually, FA block [34] also use addition based skip connections to fuse the internal input and output features.”