FFA-Net: A Feature Fusion Attention Network for Single Image Dehazing This paper proposes an end-to-end feature fusion attention network (FFA-Net) for single image dehazing. The FFA-Net architecture consists of three key components: a novel Feature Attention (FA) module that combines channel attention with pixel attention, a basic block structure consisting of local residual learning and feature attention, and an attention-based feature fusion structure. The FA module treats different features and pixels unequally, providing additional flexibility in handling different types of information. The basic block structure allows less important information to be bypassed through multiple local residual connections, allowing the main network to focus on more effective information. The attention-based feature fusion structure retains shallow layer information and passes it into deep layers, and adaptively learns different weights for different level features. The experimental results demonstrate that 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. The FFA-Net outperforms other methods in terms of image detail and color fidelity, especially in regions with thick haze and rich texture details. The FFA-Net also has a powerful advantage in the restoration of image detail and color fidelity. The FFA-Net is designed to handle the uneven distribution of haze across an image and the different weighted information of channel-wise features. The FA module combines channel attention and pixel attention mechanisms, allowing the network to focus more on important features. The basic block structure allows the network to bypass less important information, improving the network's performance. The attention-based feature fusion structure retains shallow layer information and passes it into deep layers, and adaptively learns different weights for different level features. The FFA-Net is evaluated on the RESIDE benchmark dataset and shows superior performance in terms of PSNR and SSIM metrics. The results show that the FFA-Net outperforms previous state-of-the-art methods in both quantitative and qualitative comparisons. The FFA-Net is also tested on realistic hazy images and shows strong performance in terms of image detail and color fidelity. The ablation study shows that each component of the FFA-Net plays an important role in the network's performance, especially the FFA structure. The FFA-Net is expected to solve other low-level vision tasks such as deraining, super-resolution, and denoising.FFA-Net: A Feature Fusion Attention Network for Single Image Dehazing This paper proposes an end-to-end feature fusion attention network (FFA-Net) for single image dehazing. The FFA-Net architecture consists of three key components: a novel Feature Attention (FA) module that combines channel attention with pixel attention, a basic block structure consisting of local residual learning and feature attention, and an attention-based feature fusion structure. The FA module treats different features and pixels unequally, providing additional flexibility in handling different types of information. The basic block structure allows less important information to be bypassed through multiple local residual connections, allowing the main network to focus on more effective information. The attention-based feature fusion structure retains shallow layer information and passes it into deep layers, and adaptively learns different weights for different level features. The experimental results demonstrate that 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. The FFA-Net outperforms other methods in terms of image detail and color fidelity, especially in regions with thick haze and rich texture details. The FFA-Net also has a powerful advantage in the restoration of image detail and color fidelity. The FFA-Net is designed to handle the uneven distribution of haze across an image and the different weighted information of channel-wise features. The FA module combines channel attention and pixel attention mechanisms, allowing the network to focus more on important features. The basic block structure allows the network to bypass less important information, improving the network's performance. The attention-based feature fusion structure retains shallow layer information and passes it into deep layers, and adaptively learns different weights for different level features. The FFA-Net is evaluated on the RESIDE benchmark dataset and shows superior performance in terms of PSNR and SSIM metrics. The results show that the FFA-Net outperforms previous state-of-the-art methods in both quantitative and qualitative comparisons. The FFA-Net is also tested on realistic hazy images and shows strong performance in terms of image detail and color fidelity. The ablation study shows that each component of the FFA-Net plays an important role in the network's performance, especially the FFA structure. The FFA-Net is expected to solve other low-level vision tasks such as deraining, super-resolution, and denoising.